Skip to content
Results
$28M+ Revenue GeneratedFor our clients2,140+ Keywords — Page 1Google rankings$12M+ Ad Spend ManagedAcross channels2.5M+ Signups DrivenUser acquisitions87,200+ Leads GeneratedQualified pipeline
Get proposal

WhatsApp Chatbot with Human Handoff

Build a WhatsApp chatbot your customers can use and your team can control.

We'll reply with a written proposal: the work, what we need from you and an estimated timeline.

The hidden problems

Why a chatbot that sounds good can still give old answers, do too much or fail to recover

WhatsApp automation fails when a demo is turned into a live service without clear limits on what it handles, current information, identity checks, limits on what its tools can do, protection against doing the same action twice, testing, monitoring or a handoff to a person that actually works.

  • The demo decides what gets built

    A convincing demo that only shows things going well hides the requests it can't handle, sensitive topics and different customer situations.

  • No one knows where answers come from

    Answers mix old web pages, private data and the AI's own guesses, with no owner for each source and no date when it goes out of date.

  • The chatbot has too much power

    Broad logins and unclear instructions let the chatbot take actions no one intended, or actions that can't be undone.

  • A chat message is treated as approval

    A customer's message is wrongly taken as proof of who they are, as permission, or as approval for a high-impact action.

  • The handoff loses the conversation

    The staff member who takes over gets no chat history, no idea what the customer wanted, no proof, no record of what was done and no idea what the customer was promised.

  • Sounding good becomes the measure of success

    A high reply rate hides wrong answers, failed actions, unsolved problems, slow replies, costs and harm to customers.

Exactly what's included

What our WhatsApp chatbot service can include

A WhatsApp chatbot answers your customers automatically inside WhatsApp. Softscotch takes one clearly limited customer task and turns it into a WhatsApp service your team can watch over. We agree what the chatbot may answer or do, connect it to up-to-date information and to tools that can only do what they are allowed to, test what happens when things go wrong or get risky, make sure a person can always take over, and report a smooth-sounding reply separately from a problem that was actually solved.

Eight areas of work connect your WhatsApp account setup, the customer task, the information the chatbot uses, how it keeps track of each chat, its connections to your systems, the handoff to your staff, testing and proof from the live service.

Customer requests, risks and limits

Agree which customer tasks the chatbot supports, the different customer situations, sensitive topics, what it must never do, how customers prove who they are and which decisions only a person can make.

Information the chatbot uses

List the approved information sources, who owns each one, what it covers, how up to date it is, how the chatbot looks things up, how it shows sources, privacy rules and when it should safely say "I don't know".

Conversation design

Build fixed menus or WhatsApp Flows (simple in-chat forms), memory of where each chat is up to, follow-up questions, backup options for other languages and ways to recover when a chat goes wrong.

Connections to your systems

Set strict limits on the data formats, permissions and logins each connection uses, prevent the same action running twice, check inputs, confirm actions with the customer, check the result afterwards and keep a record as proof.

Handoff to your staff

Send each chat, with its topic, urgency, language and full history, to the right named team with agreed response times, so the conversation can carry on safely.

Testing, security and re-testing

Test normal chats, unusual cases, people trying to trick the bot, privacy, prompt injection (messages that try to override the bot's instructions), out-of-date information, failed actions and handoffs to staff.

Quality and performance tracking

Report separately on whether answers are backed by your sources, where each task is up to, the results of actions, handoffs, response speed, cost, problems solved and the results you accept as real.

WhatsApp account, number and message templates

Set up or review your WhatsApp Business account and phone number, how customers agree to receive messages, any business checks Meta asks for and the message templates Meta must approve before you can start a conversation. Meta makes the final approval decisions.

Why Softscotch

Give the chatbot only the information and powers it needs

Softscotch keeps a clear trail for every chat: where the answer came from, who the customer is, what action was asked for, what your system actually did, what a person decided and what result was accepted.

What the chatbot handles is agreed in writing

The chatbot knows what it may answer, what it may ask, what it may look up, what it may do, what it should pass on and what it must refuse.

Every information source has an owner and an expiry date

Each source of facts comes with rules on what it covers, how fresh it must be, who can see it and what happens if it is missing.

Connections get the smallest possible powers

Data formats, logins and action permissions stay narrow, easy to test and easy to switch off.

People approve important decisions

Identity checks, permission, exceptions and high-impact actions are never assumed just because a chat sounded convincing.

Planning for failure comes first

Follow-up questions, saying "I don't know", trying again, undoing actions and handing over to a person are all tested before launch.

Live proof matters more than a good demo

Re-test results, live tracking, customer results and the history of past problems decide every expansion.

Service decisions

The 36 decisions behind a WhatsApp chatbot you can support

See how customer request types, business rules, permission to message, message templates (pre-approved WhatsApp messages), system connections, backup paths, handoff to people, testing and monitoring come together in a live WhatsApp chatbot.

Limits and starting point
What the chatbot will and won't doAgree which customer tasks are supported, what is not allowed, what counts as solved and who owns the other channels next to WhatsApp.
Sorting customer requests by riskGroup request types by whether they are supported, how sensitive they are, what proof they need, whether identity checks are needed, what action they lead to and whether a person must step in.
List of approved information sourcesRecord the approved sources, their owners, what they cover, who can see them, how fresh they are, where they disagree and when they are retired.
Identity, privacy and permission rulesMap how customers are identified, why each piece of data is used, what must be told to customers, how long data is kept and what permission is needed.
Staff teams and who handles escalationsName the staff teams, their hours, response times, the details they receive, who handles exceptions and who can pause the chatbot.
Defining success and safety measuresMeasure separately the quality of answers, the status of actions, problems solved, harm, value, response speed and cost.
Incoming messages and identity
Receiving and cleaning up incoming messagesCheck each incoming message for who sent it, its type, its order, duplicates, repeat deliveries and content the chatbot can't handle.
Remembering where each chat is up toDecide when a chat starts and ends, what the chatbot knows about the customer, what it remembers, when that memory expires, how clashes are handled and how it recovers.
Fixed menus and WhatsApp Flows formsUse step-by-step menus and forms for simple, limited tasks, with input checks, backup options and accessibility for all users.
FAQ answers with sourcesBase answers on approved, current information and show customers where the answer came from when that helps.
Your systems and chat status
Correct product, price and stock answersLook up current product, price and stock facts, each with an owner, a time stamp and the right region, and a safe fallback when the data isn't available.
Qualifying leads and collecting detailsAsk only the few questions that matter, check the answers and save half-finished chats so nothing is lost.
Booking and appointment connectionsCheck open slots, time zones, who the customer is, confirmations, double bookings, changes and cancellations.
Order and support case status checksConfirm who the customer is first, then share only the current order or case details they are allowed to see.
Building and delivery
Syncing contacts and chats to your CRMSave reliable records in your CRM (customer records system) of each contact, their permissions, where they came from, a chat summary, the owner and the next step.
Handling many languagesTest that the chatbot understands meaning in each language and region, copes with mixed languages and unsupported messages, and sends chats to the right person when it can't help.
Reply rules and tone of voiceKeep dated versions of required notices, limits on claims, writing style, advice the chatbot must never give and the wording used when passing a chat to a person.
When the chatbot should not answerPass the chat on when the chatbot doesn't have enough proof, isn't sure what the customer wants, can't confirm who they are, lacks permission or can't confirm an action worked.
Safety checks and testing
Asking again and recovering from errorsDeal with unclear requests, wrong inputs, timeouts and actions that only partly worked, without ever pretending something succeeded.
Identity checks and account accessKeep "who seems to be chatting" separate from a properly verified login to their account and the right to take actions on it.
Passing the full chat to a personPass on what the customer wants, the proof, their situation, what was already tried, the results of any actions and what they are expecting next.
Logging messages and their statusKeep a record of each message, its status, any retries, corrections and owner changes, with reliable time stamps.
Improvement and testing
Reviewing request types and fallbacksCompare requests that were handled, unclear, passed on or sent to staff, along with the information the chatbot used for each.
Watching resolutions and escalationsTrack tasks that were accepted as complete and what happened after staff took over, without claiming the chatbot caused those results.
Tuning speed and running costsAdjust the AI model, how information is looked up and how actions run, while keeping to clear quality and safety limits.
Sharing only approved informationGive each chat only the approved, current information that fits its purpose, and nothing more.
Measuring results and value
Clear rules for every connected actionDefine each action the chatbot can take: the exact inputs, the allowed states, input checks, protection against running twice, retries and a check of the result afterwards.
Protection against trick messages and data leaksTest content from untrusted sources, how passwords and keys are protected, attempts to pull data out, clashing instructions and keeping each chat separate.
Decisions that need a person's approvalRequire a clear decision from a person for anything sensitive, unusual or impossible to undo.
Test conversations and re-testingKeep dated sets of typical, unusual, hostile and recovery test chats, each with a clear pass mark.
Tracking quality, speed and costMonitor whether answers are backed by sources, failures, actions, handoffs to staff, response time and spending.
Rules and upkeep
Tests with hostile or tricky usersTry out manipulation, people pretending to be someone else, attempts to pull out data, unsafe advice and attempts to get around your rules.
Keeping information up to dateRefresh a source, show a warning, stop answering or retire the source once it gets too old or no longer has an owner.
Sending sensitive topics to peopleSend regulated, high-risk or professional questions, and chats with vulnerable customers, to approved staff.
No double actions, and checking resultsStop actions from happening twice and check the real result in your main system after every action.
Undo plan and record of changesPause the chatbot, remove its access, keep the evidence, put things right for affected customers and control how it restarts.
Research and implementation tools

WhatsApp chatbot tools we use, and why

The platforms we name are current examples, not fixed choices. Before we use any of them, we check your access, plan, account type, country and which features actually work for you.

Look under the hood

Seven checks that follow a customer request from start to result

Each check follows a request through the information used, the action taken and the recovery if something fails, to the final result. Each one ends with dated proof, a clear test of what counts as working, a list of exceptions and a named person who owns the service.

Check 01 · Customer requests, risks and limits

Customer requests, risks and limits

We agree which customer tasks the chatbot supports, the different customer situations, sensitive topics, what it must never do, how customers prove who they are and which decisions only a person can make.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

#1
Proof proveList customer tasks, situations and sensitive topics using real inquiry records, your service policies and the business owners responsible.
Rules defineMark each request type as answer, ask, look up, act, pass on or refuse, with its own rule for identity checks, approval and handing to staff.
Build buildLink every supported task to an approved reply path, a staff team who can take over and an agreed meaning of "solved".
When things go wrong riskTest vague, off-topic, sensitive, fake-identity and conflicting requests, and make sure the chatbot never gives itself powers it doesn't have.
Sign-off sign-offYour business, service, privacy and security owners sign off the cases the trial may handle and the cases it must pass to a person.
Next step decideRecord unsolved request types, banned actions, gaps in staff cover, who can pause the chatbot and who owns each decision before launch.
Check 02 · Information the chatbot uses

Information the chatbot uses

We list the approved information sources, who owns each one, what it covers, how up to date it is, how the chatbot looks things up, how it shows sources, privacy rules and when it should safely say "I don't know".

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveList the exact pages, catalogs, policies, records and data fields the chatbot may use, with each owner and the date it was last checked.
Rules defineSet which source wins in a clash, which customers can see what, how fresh it must be, how sources are shown, when it expires and when the chatbot should safely not answer.
Build buildLook up only the content that fits the purpose of the chat, and save which version of the source was used for each test answer.
When things go wrong riskTest missing, out-of-date, contradicting, private and tampered sources, plus failed lookups and languages the chatbot doesn't support.
Sign-off sign-offReviewers check across a realistic set of questions that answers are backed by facts, private data stays private, sources are useful and the chatbot passes on what it can't answer.
Next step decideAgree what triggers a refresh, how retired sources are handled, how disputed answers are dealt with and who can approve a change to the information.
Check 03 · Conversation design

Conversation design

We build fixed menus or WhatsApp Flows (simple in-chat forms), memory of where each chat is up to, follow-up questions, backup options for other languages and ways to recover when a chat goes wrong.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveMap how customers start a chat, the type of message, when a chat starts and ends, the customer's situation and whether they use a form or type freely.
Rules defineAgree follow-up questions, checks on what customers type, what the chatbot may remember, language and region behavior, timeouts and a safe way back to a person.
Build buildCarry only approved details from one message to the next, and make every step, correction and timeout visible in the records.
When things go wrong riskTest messages delivered twice or out of order, half-finished forms, mixed languages, wrong entries and chats that have timed out.
Sign-off sign-offConfirm customers can understand the choices, fix what they typed, leave the automated chat and reach a real person with the chat history kept.
Next step decideRecord confusing steps, accessibility limits, recovery problems, changes to what is remembered and who owns each fix going live.
Check 04 · Connections to your systems

Connections to your systems

We set strict limits on the data formats, permissions and logins each connection uses, prevent the same action running twice, check inputs, confirm actions with the customer, check the result afterwards and keep a record as proof.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveConfirm each API (a secure link between software systems) needed, who owns its login, the type of data involved, what it changes and which system holds the true record afterwards.
Rules defineLimit each connection's data format, inputs, identity checks, permissions, confirmation step, retries, protection against running twice and how actions are undone.
Build buildLog separately the action suggested, the checked input, the approval, the request number, the reply from the other system and the final check of the real result.
When things go wrong riskTest actions that are not allowed, run twice, time out, only partly finish or falsely report success, without causing the same effect twice.
Sign-off sign-offTechnical and business owners repeat the allowed actions themselves and confirm that blocked actions leave the customer and your systems safe.
Next step decideList open connection problems, when logins must be changed, limits set by other providers, steps to undo changes and who owns any new actions added later.
Check 05 · Handoff to your staff

Handoff to your staff

We send each chat, with its topic, urgency, language and full history, to the right named team with agreed response times, so the conversation can carry on safely.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveName the staff teams, their hours, languages, urgency rules, response times, the details passed to them and the people who can take over.
Rules defineAgree when the chatbot must hand over, what the customer is told and which chat history, proof and action status go with the chat.
Build buildPass the live chat on with the customer's request, their situation, failed attempts and the next decision still clearly visible.
When things go wrong riskTest what happens when no staff are free, queues are full, topics are sensitive, chats are passed around, history is lost or the chatbot must be paused urgently.
Sign-off sign-offYour service managers confirm chats go to the right place with full history, customers know what to expect and the chat carries on safely after a person takes over.
Next step decideTrack hours with no staff cover, queue failures, chats left unsolved, what depends on staffing levels and who owns each fix to the recovery process.
Check 06 · Testing, security and re-testing

Testing, security and re-testing

We test normal chats, unusual cases, people trying to trick the bot, privacy, prompt injection (messages that try to override the bot's instructions), out-of-date information, failed actions and handoffs to staff.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveBuild a dated set of test chats: normal, unusual, hostile, privacy, unsafe actions, out-of-date sources and recovery.
Rules defineSet pass marks for staying within scope, factual accuracy, permission checks, privacy, action status, handoff to staff and harm to customers.
Build buildRun the same approved test chats against the exact instructions, sources, connections, AI model, Flow and rules planned for launch.
When things go wrong riskTry prompt injection, requests for passwords or secrets, fake identities, misuse of connections, tampered sources, unsafe advice and systems that are down.
Sign-off sign-offReviewers look at every high-risk failure, plus a fair sample of the answers, actions and handoffs to staff that passed.
Next step decideShare the failed tests, how serious each is, the fix, proof from re-testing, the risk that remains and the person allowed to launch or pause the chatbot.
Check 07 · Quality and performance tracking

Quality and performance tracking

We report separately on whether answers are backed by your sources, where each task is up to, the results of actions, handoffs, response speed, cost, problems solved and the results you accept as real.

Each check ends with dated proof, the cases you approved, the limits still open and a named person who owns the next step.

Proof proveSet fixed IDs and clear "out of how many" counts for request types, answers, actions, handoffs, problems solved, response speed, cost and incidents.
Rules defineKeep separate: whether a source backs the answer, the fact check, the action tried, what your main system shows, a person closing the case and the value you accept.
Build buildLink chatbot records to service results only through approved IDs, time windows and privacy rules.
When things go wrong riskLook into missing records, late updates, duplicate statuses, silent failures, sudden cost jumps and records that disagree about whether a case was closed.
Sign-off sign-offCompare the trial with the agreed quality and safety pass marks before adding more customers, request types, sources or actions.
Next step decideHand over dashboards, alerts, spot-check reviews, how often information is refreshed, a plan for handling problems and the named person who decides the next update.
Your type of business

The approach changes with your type of business

The plan adapts to your customers, your level of risk, your data, how often things change and how you make money, while the chatbot stays within the same agreed limits.

Local

Local services

For local service businesses, a WhatsApp chatbot has to allow for your locations, the areas you serve, availability, phone calls, bookings and who replies to each customer.

FocusReceiving and cleaning up incoming messages, remembering where each chat is up to, fixed menus and WhatsApp Flows forms
MeasureRequest types and coverage, answer accuracy and sources, actions taken and their status
Ecommerce

Online stores and direct-to-consumer brands

For online stores and brands that sell direct, a WhatsApp chatbot has to allow for correct product details, stock levels, promotion terms, checkout steps, returns and profit margin.

FocusCorrect product, price and stock answers, qualifying leads and collecting details, booking and appointment connections
MeasureAnswer accuracy and sources, actions taken and their status, handoffs to staff and service
Software

Software companies (SaaS)

For software companies, a WhatsApp chatbot has to allow for each customer's use case, proof about your product, free trial or demo steps, connections to other tools and getting new users started.

FocusSyncing contacts and chats to your CRM, handling many languages, reply rules and tone of voice
MeasureActions taken and their status, handoffs to staff and service, speed and cost
B2B

Business-to-business (B2B) and professional services

For B2B and professional services, a WhatsApp chatbot has to allow for several people being involved in buying, proof of expertise, lead qualifying, handoff to sales and the long gap before a deal closes.

FocusAsking again and recovering from errors, identity checks and account access, passing the full chat to a person
MeasureHandoffs to staff and service, speed and cost, resolved issues and business results
Health

Healthcare and wellness

For healthcare and wellness, a WhatsApp chatbot has to allow for accuracy, privacy, who is eligible for care, appointment status and medical advice it must never give.

FocusReviewing request types and fallbacks, watching resolutions and escalations, tuning speed and running costs
MeasureRequest types and coverage, answer accuracy and sources, actions taken and their status
Regulated

Finance, insurance and legal

For finance, insurance and legal firms, a WhatsApp chatbot has to allow for the laws in each region, approvals, required notices, sensitive data and terms that change.

FocusClear rules for every connected action, protection against trick messages and data leaks, decisions that need a person's approval
MeasureAnswer accuracy and sources, actions taken and their status, handoffs to staff and service
Education

Education and training

For education and training, a WhatsApp chatbot has to allow for course and program facts, dates, who can apply, enrollment steps, what students can expect to achieve and accessibility for every learner.

FocusTests with hostile or tricky users, keeping information up to date, sending sensitive topics to people
MeasureActions taken and their status, handoffs to staff and service, speed and cost
Media

Publishers and media

For publishers and media, a WhatsApp chatbot has to allow for who wrote what, content rights, subscriptions, content that changes very quickly and original reporting or research.

FocusWhat the chatbot will and won't do, sorting customer requests by risk, list of approved information sources
MeasureHandoffs to staff and service, speed and cost, resolved issues and business results
Marketplaces

Marketplaces and directories

For marketplaces and directories, a WhatsApp chatbot has to allow for serving both buyers and sellers, identity checks, listings, data quality and actions that might happen twice.

FocusReceiving and cleaning up incoming messages, remembering where each chat is up to, fixed menus and WhatsApp Flows forms
MeasureRequest types and coverage, answer accuracy and sources, actions taken and their status
Travel

Travel and hospitality

For travel and hospitality, a WhatsApp chatbot has to allow for busy and quiet seasons, location, availability, the value of each booking, cancellations and the rules in your booking policies.

FocusCorrect product, price and stock answers, qualifying leads and collecting details, booking and appointment connections
MeasureAnswer accuracy and sources, actions taken and their status, handoffs to staff and service
Industrial

Industrial and manufacturing

For industrial and manufacturing firms, a WhatsApp chatbot has to allow for technical specs, what works with what, distributors, long buying decisions and good-quality inquiries.

FocusSyncing contacts and chats to your CRM, handling many languages, reply rules and tone of voice
MeasureActions taken and their status, handoffs to staff and service, speed and cost
Membership

Memberships and communities

For memberships and communities, a WhatsApp chatbot has to allow for each member's access level, the ongoing value they get, moderation, renewals and support for members.

FocusAsking again and recovering from errors, identity checks and account access, passing the full chat to a person
MeasureHandoffs to staff and service, speed and cost, resolved issues and business results
Clear ownership

Who does what: product, security, customer service and legal each have their own owner

We spell out what we do ourselves, what we coordinate with others and which decisions stay with you.

We handle

The clearly limited WhatsApp chatbot itself, plus the proof and setup work behind it.

  • Customer requests, risks and limits
  • Information the chatbot uses
  • Conversation design
  • Connections to your systems
  • Handoff to your staff
  • Testing, security and re-testing
We coordinate

The outside teams and inputs needed to launch safely and read the results correctly.

  • Owners of your WhatsApp account and inbox
  • Owners of your products and information
  • Your app and software teams
  • Your login and security teams
  • Your customer service teams
  • Legal and privacy reviewers
You keep

The business, legal and launch decisions that stay with you.

  • Which requests are supported, and your product rules
  • The true facts and approval of all information
  • Customer identity and permission rules
  • Control of connections and logins
  • Approval of sensitive decisions and exceptions
  • Going live, pausing and handling problems when they happen
How it fits together

How a WhatsApp chatbot stays manageable: nine practical records

Your team can see what the chatbot covers, where its answers come from, how customers are identified, what actions it took, how it was tested, how chats were handed to staff, what failed in live use and what the next decision about changes is.

WhatsApp Chatbot and AI Agent

01
Signed list of what the chatbot may do

One approved record names the customer tasks the chatbot may handle, the actions it may never take and the people responsible for exceptions.

02
Identity and data map

For each stage of a chat, it shows what is known about who the customer is, what data may be used, what must be told to the customer and when a stronger identity check is needed.

03
Message log

Fixed IDs keep track of messages that were received, duplicated, out of order, sent again, not supported or corrected.

04
List of approved information

Every source the chatbot can use has an owner, a clear scope, an access rule, a review date, a rule for clashes and a safe way to retire it when it goes out of date.

05
Rules for each connection

Each connection to your systems has fixed input formats, logins with limited powers, identity and confirmation steps, protection against repeat actions and a check of the real result.

06
Reply and refusal guide

Approved wording covers required notices, claims, sensitive topics, uncertainty, refusing a request, correcting a mistake and what the customer is promised when handed to a person.

07
Test results report

Normal, unusual, hostile and recovery test chats show the exact version tested, the pass mark, any failure and the reviewer's decision.

08
Handoff guide for staff

Named staff teams receive the chat history, what the customer wants, the proof, the action status and what the customer expects, with staff hours and service limits clearly shown.

09
Trial running pack

Dashboards, alerts, spot-check reviews, information refreshes, emergency pause, undo steps and the right to expand all stay with named owners.

The connected growth engine

Product

Owns which requests are supported, what customers are promised, policies, what data means and what counts as solved.

Knowledge

Owns information sources, who can see them, how fresh they are, how they are shown, gaps and retiring old sources.

Engineering

Owns incoming message handling, chat memory, connections, data formats, protection against repeat actions, monitoring and undo steps.

Security

Owns identity, passwords and keys, giving the smallest possible access, protection against trick messages and handling security problems.

Service

Owns staff teams, response times, escalations, recovery and putting things right for customers.

Rules and upkeep

Owns privacy, sensitive actions, professional limits, claims and the final go-live decision.

One connected scope. WhatsApp Chatbot and AI Agent is coordinated with the strategy, creative, experience, data and reporting it depends on, so responsibilities and handoffs stay visible in one written plan.

What to expect

From one clearly limited customer task to a careful trial run

Finding out what you need and designing the first limited version usually takes 1 to 2 weeks once we have everything we need from you. A careful trial run (pilot) with real customers usually takes 4 to 10 weeks, depending on your information sources, system connections, identity checks, review time, risk level and whether your staff team is ready to take over chats.

WhatsApp chatbot and AI agent, from first review to a limited trial

Week 1

Agree the limits

Approve the customer task for the WhatsApp chatbot, what is left out, the markets, the owners and who makes decisions.

KPI Scope
Weeks 1–2

Customer requests, risks and limits

Agree which customer tasks the chatbot supports, customer situations, sensitive topics, what it must never do, how customers prove who they are and which decisions only a person can make.

KPI Review
Weeks 2–4

Connections to your systems

Set strict limits on data formats, permissions and logins, prevent repeat actions, check inputs, confirm actions, check results afterwards and keep proof.

KPI Build
Weeks 4–6

Test what happens when things fail

Test requests it doesn't support, unclear messages, old or tampered sources, prompt injection (trick messages), unknown identity, actions that aren't allowed, repeat actions, partial failures, timeouts, no staff available, privacy requests and undoing changes.

KPI Failure tests
Weeks 6–10

Launch approved work

Launch only the request types and connections that passed testing, with current sources, the smallest possible access, required approvals, enough staff to take over, full records, monitoring, a pause switch and undo steps in place.

KPI Go live
Months 3–4

Hand over and keep it running

Assign who refreshes information, how often tests are re-run, when logins are changed, connection reviews, staff cover, handling problems, putting things right for customers and who decides on expanding.

KPI Run

Our full growth plan (all services together), month by month

Month 1

Foundation and review

We review your whole website, where you rank and your competitors, agree how the site should be organized, and set up hosting, analytics (visitor tracking) and your GoHighLevel lead system. The first agreed pages start going live.

KPI pages live
Month 2

Build and launch

The site is built or rebuilt. Main, service and industry pages go live with on-page SEO, website labels and technical fixes. Your Google Business Profile is improved and lead follow-up automations are switched on.

KPI pages in Google
Month 3

Early rankings, paid ads start

More pages built from templates (programmatic pages) start going live, paid ads and email follow-ups are switched on for early leads, and Google adds your pages to its results as the first rankings usually appear.

KPI search terms ranking
Month 4

Momentum builds

Rankings often start climbing, visits from free search begin to rise, ad spending is tuned to the return it brings (ROAS, return on ad spend), and groups of related articles and links from other sites build your reputation with Google.

KPI visits from search
Month 5

Results start to show

Main search terms can move closer to page one, the first leads from free search often arrive, AI tools may begin to mention you, and testing improvements to your pages (CRO, conversion rate optimization) helps more visitors get in touch.

KPI leads from search
Month 6

Growth builds on itself

Pages that are working are protected and improved, template pages bring in visitors searching for very specific terms, and SEO, ads, email and automation work together to bring steadier leads and sales. Timing varies by market and competition, and no ranking is guaranteed.

KPI leads and sales
How we compare

A chatbot is only as reliable as the limits and backup plans around it

This comparison is about proof and control, not claims that we are better than everyone else, which no one can prove.

The risk of using a different agency for everything
  • Starts with a vague promise to automate all customer support
  • Treats a convincing demo as proof it's ready for real customers
  • Lets the AI's memory and undated files answer customers
  • Gives one system login much more power than the task needs
  • Treats "who seems to be chatting" as a verified login to their account
  • Hides failed or half-finished actions behind a friendly reply
  • Tests only expected questions and chats where everything goes well
  • Passes chats to staff without the proof or action status they need
  • Reports message and reply counts without showing problems actually solved
  • Promises accuracy, fewer staff chats or savings it can't control
vs
Softscotch
  • Starts with one clearly limited customer task and named exceptions
  • Checks the live WhatsApp setup, its features, your systems and how many staff can take over
  • Uses approved sources with owners, scope, freshness rules and rules for when not to answer
  • Limits each connection by identity, data format, permission, confirmation and a check of the result
  • Keeps what was said in the chat separate from verified permission to act
  • Records actions that were suggested, tried, successful, failed and undone
  • Tests normal, unusual, hostile, privacy and recovery cases
  • Hands over the chat history, proof, attempts and the next decision needed
  • Separates reply quality, action status, staff closing a case and problems actually solved
  • Ends with named owners for launching, pausing, upkeep and expanding
The numbers

Results from Softscotch client work.

These totals come from work for Softscotch clients across search, paid ads, websites and lead generation, as published on softscotch.com. Past results don't guarantee future results.

$12M+Ad spend managed for clients
87,200+Leads generated for clients
2,140+Search terms ranked on Google's first page
$28M+Revenue generated for clients
Client case studies

All six Softscotch client stories.

Every Quick Story currently published on Softscotch.com, covering education, ecommerce, healthcare and SaaS. Stock photography represents each client industry; the figures are reported historical results and do not guarantee future performance.

Education consultant speaking with a prospective student Education Ads + local SEO
+183%leads

183% more leads for an education consultancy

WordPressQuora adsLocal SEO
The challenge

An education consulting business in India needed to build its digital presence and lead flow from scratch.

What changed
  • Built a new WordPress website
  • Used Quora ads to grow daily lead volume
  • Added Local SEO and community engagement
+183% leads25+ leads/day-38% bounce rate+273% new users
Bright modern home interior representing the home-solutions client Home solutions Shopify + SEO
+82%Shopify sales

Home-solutions ecommerce growth

ShopifyOrganic searchAmazon ads
The challenge

A home-solutions business was dependent on business-to-business (B2B) marketplaces, losing sales and recovering from a Google penalty.

What changed
  • Resolved the Google penalty
  • Revamped the website and online store
  • Improved organic visibility and Amazon campaign performance
+82% Shopify sales+147% organic+23% Amazon ROAS
Mother and baby representing the lactation-consultant client Health consultant $20/day
5/daybooked calls

Five booked calls a day for a health consultant

Facebook adsGoogle AdsLocal SEO
The challenge

A Houston lactation consultant needed to recover a malware-damaged website and rebuild local patient acquisition.

What changed
  • Rebuilt the hacked site and migrated it to secure hosting
  • Ran targeted Facebook and Google campaigns
  • Reached top-three local Google positions
5 calls/day$20 daily spend+87% local traffic+34% new users
Ecommerce owner preparing customer orders beside a laptop Ecommerce 10 months
+315%organic traffic

315% organic growth for an ecommerce site

Content strategyOn-page SEO
The result

Softscotch reports a 315% increase in ecommerce organic traffic over 10 months through focused on-page SEO.

What changed
  • Optimized product and collection pages
  • Improved the links between pages and the page addresses (URLs)
  • Made the site easier for Google to explore and better matched to what people search for
+315% organic10 monthsProductsCollections
SaaS team reviewing campaign data around a laptop SaaS platform Cold outreach
16%reply rate

9% close rate for a transcription SaaS

ProspectingSMTP
The challenge

A transcription platform needed a qualified prospect list and a reliable email-sending setup for outreach we managed.

What changed
  • Built a comprehensive prospecting list
  • Managed the cold outreach campaigns
  • Set up reliable email-sending servers (SMTP)
9% close rate16% reply rateSaaSManaged outreach
Healthcare professional consulting with a patient Healthcare 6 months
+182%organic traffic

182% organic growth for a healthcare clinic

HealthcareOn-page SEOOff-page SEO
The challenge

A clinic in a competitive location needed to build organic search visibility from the ground up.

What changed
  • Built the clinic's foundation for unpaid (organic) Google search
  • Combined on-page and off-page SEO
  • Reached top-five positions in Google Search
+182% organic6 monthsTop 5 Google
01 / 06
In their words

Teams that stopped leaving clicks on the table.

★★★★★

"The Softscotch team has been a solid partner as our business evolved over the last five years. Unlike most agencies, they have never charged us for work that falls outside the agreement. That is what sets them apart."

IM
Imtiyaz MominHairareus
★★★★★

"Without their knowledge, we would be in a very different place than we are now. Whenever they find something that can be improved, they implement it quickly. I could not be happier with their performance and response time."

AP
Aaron PaulsonBaby Pavilion
★★★★★

"Softscotch goes above and beyond the scope. The time, effort and dedication they put in is like having your own in-house marketing team. They are not just another marketing company."

TM
Tabassum MithaniTRS Care
★★★★★

"We are extremely happy with Softscotch, such an attentive team. No project is too small, and they give their 100%. We are extremely satisfied with their work."

SH
Safiya HafiziSimply Lactation
★★★★★

"Softscotch is a great team to work with: transparent, reliable and genuinely invested in the growth of your company. They are good partners to have by your side."

MJ
Malvika JainCyclop
★★★★★

"The Softscotch team has been a solid partner as our business evolved over the last five years. Unlike most agencies, they have never charged us for work that falls outside the agreement. That is what sets them apart."

IM
Imtiyaz MominHairareus
★★★★★

"Without their knowledge, we would be in a very different place than we are now. Whenever they find something that can be improved, they implement it quickly. I could not be happier with their performance and response time."

AP
Aaron PaulsonBaby Pavilion
★★★★★

"Softscotch goes above and beyond the scope. The time, effort and dedication they put in is like having your own in-house marketing team. They are not just another marketing company."

TM
Tabassum MithaniTRS Care
★★★★★

"We are extremely happy with Softscotch, such an attentive team. No project is too small, and they give their 100%. We are extremely satisfied with their work."

SH
Safiya HafiziSimply Lactation
No black box

You'll always know exactly what we changed.

Plenty of agencies go quiet after the audit and you never quite know what you're paying for. We do the opposite. You get a live view of every change and what it moved, in plain language you can act on.

Your delivery, at a glanceSample
Changes shipped12
Pages reviewed28
StartMidpointNow
Delivery activity✓ Up to date
HTML mockupExplore mockup ↗

Live client dashboard

Every change, score and ranking in one place you can open any time, not a monthly PDF you wait around for.

This week, explainedSample
✓Titles updated
Ready for your review
SSSoftscotch weekly update
0:00
2:00
HTML mockupExplore mockup ↗

Weekly video summary

A short screen-recorded video (Loom) each week walking you through exactly what shipped and what it moved, so nothing is a mystery.

The change is clearSample

Before

Titles30 / 100
Structure42 / 100
Content28 / 100

After

Titles84 / 100
Structure79 / 100
Content90 / 100
Same page. Reviewed changes.
HTML mockupExplore mockup ↗

Before/after score per page

See the exact lift on every page we touch, not a vague "it's improving" you have to take on faith.

Your weekly reportSample
01
What shipped

Approved page updates are live.

02
What we learned

Product details need a review.

03
What happens next

Confirm the next priority pages.

HTML mockupExplore mockup ↗

Plain-English reporting

Clicks, rankings and revenue, explained in words you can act on, instead of a wall of charts about how often you were seen (impression charts.

Your data. Your view.Sample
Organic search
StartMidpointNow
Sample report preview↗
HTML mockupExplore mockup ↗

Reports on demand

Ask for any cut of the data whenever you need it and get it, no waiting for the end of the month.

Every step, visibleSample
  1. ✓
    Proposed

    Softscotch prepares the change.

  2. ✓
    Approved

    Your team gives the go-ahead.

  3. ✓
    Published

    Checked live and recorded.

HTML mockupExplore mockup ↗

Ethical, transparent, always on

One accountable team, strictly white-hat (only methods search engines and platforms allow), no shortcuts and no surprises. Support whenever you need it.

One plan, full-stack access

One Softscotch plan gives you access to the full service stack, coordinated around your priorities.

You do not need a separate agency or retainer for every channel. Your plan can draw from search, paid media, web, ecommerce, outreach, automation, lead generation, messaging, content, creative and operations. The agreed scope sets what is active, in what order and at what capacity.

Pricing approach

One price for everything, not a stack of separate retainers.

You get one price for your whole marketing, not a separate retainer for every service. SEO, paid ads, your website, automations, cold email and newsletters, we run all of it as a single scope, so you get the most result for what you pay. Agencies sell each of these on its own, so the invoice climbs whether or not that line drives a single lead.

Bought separately, service by service

A different agency or retainer for each service. The bill stacks up fast, and keeps climbing whether or not each one moves a single lead.

SEO (technical, on-page, content)$2,000–3,500/mo
Local SEO+$750–1,500/mo
Programmatic SEO+$1,000–2,500/mo
CRO & landing pages+$2,000–3,500/mo
Paid ads (Google & Meta)+$2,500–4,000/mo
Marketing automations+$1,500–3,000/mo
Cold email outreach+$1,500–2,500/mo
Email marketing & newsletters+$750–1,500/mo
Website management & care+$750–1,500/mo
WordPress build & setup+$2,500–6,000 once
10 separate services$12,750–23,500/mo

Plus the one-off build, and a fresh contract, kickoff and onboarding delay for every new service you add.

All of it, one scope

We run your whole marketing as one engagement, so you get the most result for what you pay, not a pile of disconnected invoices.

  • SEO: technical, on-page, content, local and AI search
  • Template-built pages (programmatic SEO), links from other websites (backlinks) and authority
  • CRO and landing pages
  • Paid ads across Google and Meta
  • Automations, cold email and newsletters
  • WordPress build, setup and ongoing care
  • One team, one roadmap and one point of contact
  • Reporting tied to qualified leads and revenue
Everything above. One flat monthly fee.A fraction of the cost of the stack on the left
Plus free access to our own platforms
Post InstantlyVerified WidgetsTrending FunnelsBrands for CreatorsDarnItSEOGoHighLevel

Priced to your site after the audit. No sticker price, and no per-service creep.

What shapes your quote
Site sizePage count, templates and how much of the site actually needs work.
Market competitionHow hard the terms you want to win are, and who you are up against.
Content needsHow much new writing, rewriting and production the plan calls for.
Build complexityYour website platform (CMS, the system you edit your site in), your other tools and how involved the changes are to make.
Final pricing comes after the audit, once we have seen the four things above on your actual site. You get a written scope and a fixed quote before any work begins: no lock-in, no surprise line items, and no upfront number that ignores your market.
Request a custom scope
Questions, answered

WhatsApp chatbot questions to ask before you buy

Our answers are honest about the limits set by WhatsApp, the law, your data and what results anyone can promise, before any work begins.

Filter
What is included in your WhatsApp chatbot and AI agent service?

Your proposal can include a map of customer requests and risks, WhatsApp account, number and message template setup, a list of approved information sources, conversation and WhatsApp Flows design (simple in-chat forms), connections to your systems with limited permissions, identity check rules, handoff to your staff, testing for normal use and security, proof from a trial launch and a handover for running it day to day. The signed scope names the exact customer tasks, systems and anything left out.

When is a WhatsApp chatbot a good fit for my business?

It is a good fit when you can name a customer task that happens again and again, trusted sources of information, safe limits for system connections and people who handle the exceptions. If your information or your staff team isn't ready yet, the first project should fix that or start smaller.

What access do you need to build a WhatsApp chatbot?

We need your live WhatsApp account and inbox setup, a realistic sample of past chats, the requests the chatbot should and shouldn't handle, approved information sources, customer situations, technical details of your systems' APIs (secure links between software), test accounts, identity and permission rules, your staff teams, the people who handle problems and secure access to the systems included.

Which WhatsApp platform requirements do you check?

We check whether you use the WhatsApp Business app or the WhatsApp Business Platform, whether your account and country qualify, which message types and Flows features you can use, how incoming messages reach your systems, what outside providers are involved and whether any of Meta's own AI features for businesses are available to you. A product page or an account label is not proof that a feature actually works for you.

How do permission and message templates affect a WhatsApp chatbot?

Only information that fits the purpose of the chat reaches the chatbot. An open chat does not prove who the customer is, does not mean they agreed to marketing and does not give permission for a sensitive action. So each use of data, each type of message and each action has its own rule.

What proof do you keep before changing the chatbot?

We keep the signed list of request types, source versions, the information looked up, chat status, the details of each action and its real result, confirmations, handoffs to staff, reviewer decisions, response speed, cost, problems and the results you accepted.

How do you measure a WhatsApp chatbot?

Our reports show separately supported and unsupported requests, whether answers are backed by sources, fact checks, suggested and completed actions, handoffs to staff, response speed, cost, problems solved for customers and the business results you accept. Each number shows what it is counted out of, the time period, what was left out and how far the evidence goes.

Do delivered messages, faster pages or opened chats prove business value?

No. A delivered message, a smooth answer, information found and a request accepted by your system are all steps along the way. Whether the answer was right, whether the action was allowed, whether it was truly completed, whether the customer's problem was solved and whether it made money each need their own proof.

How do you keep everyday tracking separate from proper experiments?

Test chats and live tracking help us decide whether to launch, but they don't prove on their own that the chatbot caused a business result. To claim that, you need a fair test with a comparison group, a stable result to measure and honest notes on side effects, sample size and learning limits.

Do you build the approved WhatsApp chatbot work?

Yes, within the itemized scope. Softscotch can design and build the approved conversations, information setup, WhatsApp Flows, system connections, tests and monitoring, then check the careful trial launch. Big product changes, fixing your source content and providing customer service staff stay separate unless they are included.

Where do people stay in control of the WhatsApp chatbot?

Your team approves which tasks are supported, the true facts, how data is used, identity rules, what connections may do, sensitive or permanent actions, what customers are promised, exceptions handled by people, going live, emergency pauses, putting things right and any expansion.

Can you guarantee results from a WhatsApp chatbot?

No. Softscotch is responsible for the agreed limits, the build, the tests and the reports. We can't guarantee that WhatsApp will approve your account or features, that every answer will be correct, that it is safe to use outside the signed scope, how many chats it handles without staff, that every problem is solved, or any leads, sales, savings or return.

How long does the first WhatsApp chatbot project take?

Finding out what you need and designing the first limited version usually takes 1 to 2 weeks once we have everything we need from you. A careful trial run usually takes 4 to 10 weeks. The real plan depends on the quality of your information, system connections, identity checks, risk, review time and whether your staff team is ready to take over chats.

How is a WhatsApp chatbot priced?

The price depends on the customer tasks, information sources, languages, WhatsApp Flows, system connections, identity and security needs, how deep the testing goes, the handoff to staff, monitoring and upkeep. Fees from WhatsApp, other providers, AI models, hosting and software, plus fixing content and staffing customer service, are separate unless listed in your proposal.

Does Softscotch give legal approval for the WhatsApp chatbot?

No. Softscotch can build the documented WhatsApp, permission, privacy, security, accessibility and claim controls, and give you proof that they are in place. Your organization and qualified advisers keep the final say on what the law means and whether the chatbot can launch in each market.

What work is not part of the WhatsApp chatbot service?

Unless listed in your proposal, the scope does not include bulk broadcast campaigns, paid ads, customer service staff, wider customer lifecycle marketing (messages across the whole customer journey), fixing your source content, major behind-the-scenes product work or AI agent platforms that cover many channels.

How do you handle personal data and logins for the WhatsApp chatbot?

The access plan names the people and systems involved, gives only the permissions needed, uses approved ways to store passwords and keys, limits how long data is kept and records when access is removed. Your organization approves who does what with data, the legal basis for using it, supplier terms, limits on sensitive data and the logins used in the live system.

What happens after the WhatsApp chatbot first goes live?

The handover records the launched instructions, sources, WhatsApp Flows, connections and rules; the tests that passed and failed; live alerts; what triggers an information refresh; staff teams; problems; risks still open; and the named person who decides the next update, pause or expansion.

Related services and industries