Customers want fast answers and a human when it matters.
Faster, consistent replies that are checked before they go, with a person always within reach.
Service leaders are under pressure to use AI, while customers insist on being able to reach a person. GenWA Automate drafts and prepares replies inside your rules, GenWA Guard checks them, and the record shows who approved what.
of customers say companies that use generative AI for customer service must provide access to a human agent.
Gartner customer survey, August 2026of customer service leaders report pressure from executives to implement AI in 2026.
Gartner survey of service leaders, February 2026rise in AI spending by service leaders, against a 2% rise in overall service budgets (survey of 199 leaders).
Gartner, August 2026of consumers report frustration with AI agents, up from 54% six months earlier (US, UK and Canada).
AnswerConnect, May 2026 (vendor survey)You are asked to do more with AI, without losing the customers you already have.
of customers say a human must always be reachable when generative AI handles service.
Gartner, August 2026In 2024 a tribunal held an airline to a refund policy its chatbot invented and rejected the argument that the chatbot was a separate legal entity (Moffatt v Air Canada, 2024 BCCRT 149). What it costs you: the company is bound by what the bot says.
Gartner found AI spending by service leaders up 38% while service budgets rose 2%. What it costs you: pressure to show results from tools you cannot yet measure.
Quality teams read a small share of conversations. What it costs you: the conversations that go wrong are the ones nobody read.
Frustration with AI agents is rising and most people prefer to speak to a person (AnswerConnect, 2026, vendor survey). What it costs you: complaints, churn and a damaged name.
Different people and tools give different answers to the same question. What it costs you: inconsistent outcomes that are hard to defend as fair.
Rebuilding who said what, and why, takes senior time. What it costs you: slow, costly complaint handling.
Customers do not mind AI. They mind being stuck with it.
The research points one way: people accept automation that solves their problem quickly and resent automation that blocks them from a person. So the design question is not how much to automate; it is where AI prepares, where a rule checks and where a person takes over. Let agents draft and prepare replies from approved sources, check every reply against your rules, send anything unusual or sensitive to a named person, and keep a record of each step. The Competition and Markets Authority has said consumer law applies whether the customer is served by people or AI.
Control applies to AI used through GenWA. Credit, hiring and medical matters stay with people, not AI.
Six ways to speed up replies without losing control of what is said.
Automate does the preparation. Guard checks it. Evidence keeps the record. Advisers, human and agentic, shape it to your team.
Comms drafts letters, emails and messages, checks them against your rules and routes them to a person to sign off.
About CommsHelp for advisers mid-conversationContact gives advisers live help during a call on Amazon Connect, the cloud contact centre service from Amazon.
About ContactRules for what may be saidRules set what AI may say and where a person must take over, such as refunds, complaints and vulnerable customers.
About RulesCustomer communications reviewedGuard analyses customer messages and reports on anything that does not align with your policies and the external rules you select.
About GuardA record of every reply and every exceptionThe Log and the Exception Log record each reply, block, challenge and sign-off with the reason.
About the Exception LogThe customer service toolkitRules, Data Protection, the Log, the Exception Log, Contact and Comms, with Skill Teams for complaints (being built). Sold as part of the GenWA Toolkit. We confirm in writing what is available today.
See the GenWA ToolkitAnalyse, review, flag, report: service leaders see only what needs a decision.
Replies are read against your rules as they are prepared. What aligns can go to the customer, what does not is blocked or sent to a named person, and the pattern is reported back so you can fix causes, not only symptoms.
Animated flow showing analysis, review, flagging and reporting of customer replies
- 1Analysis
AI use is read against your rules, policies and chosen regulations.
- 2Review
Each finding is checked against the evidence behind it.
- 3Flag
Anything outside the rules is raised to a named owner.
- 4Report
What was found and what was decided is recorded and dated.
- Finding the right answer in approved sources.
- Drafting a reply for an agent to send or edit.
- Summarising a conversation for the next agent.
- Refunds, compensation and exceptions to policy.
- Complaints, bereavement and vulnerable customers.
- Any decision on credit, hiring or medical matters.
Customers judge you on how well you handled the hard moment. The routine moments are where AI helps most.
Three sectors where customer service carries the highest stakes.
Pick the sector closest to yours.
Peak trading without temporary staff
Peak seasons multiply contact volumes, and consumer law applies to every reply. Comms prepares order and returns replies within your rules, and agents take over refunds and exceptions.
See retail and consumerComplaints and customers who may be vulnerable
Complaint handling has fixed timescales and firms must show good customer outcomes. Guard checks replies and routes sensitive cases to people, and the Exception Log shows how each was handled.
See financial servicesClaims updates and policy questions
Customers expect clear, consistent updates on claims. Comms drafts updates from approved wording, and people decide on claims. The Log supports a complaint review.
See insuranceThree routes for a customer service team, from a plan to a supervised build.
Advisers are human and agentic. There is a fourth route, the Accelerator, for founder-led businesses: places are by application, and payment is by equity, project, retainer or licence.
Our advisers, human and agentic, review where AI helps your team today, set the rules for what may be said, and choose the first reply type to prepare.
See advisoryProductsComms and Contact run with your rules and a record behind every job in your AWS or Azure account, or hosted by GenWA on AWS as agreed, and can be shown in a demo.
See CommsIncubatorWe build and run a customer service function with you in an agreed engagement in a controlled environment. Named people approve, and you own the prompts, workflows and record.
See the IncubatorEstimate the hours one defined job could return to your team.
Enter how many times a job is done and how long it takes today. Treat the result as an indicative estimate built from your figures, not as a promise.
Time released calculator
Indicative, from the numbers you enter
- Time is released only for tasks handed over in full. Checking the agent's work is not counted.
- Approval of consequential actions stays with a named person.
Items marked assumption can be edited. No figures are stored or sent.
How this is worked out
Hours a week handed over = tasks x minutes each / 60 x share suitable. Hours released a month = hours a week x working weeks in a month.
Three colleagues who share the outcome, and three functions to bring alongside.
The people who own operations, risk and cost, and the teams that handle the same customers.
Owns the process, the volumes and the handover to people.
For the chief operating officerCompliance and General CounselOwns the rules on what may be said and the record.
For Compliance and General CounselChief financial officerOwns the cost of service and the case for change.
For the chief financial officerContact centresHandle the calls that need live help.
AI for contact centresCompliance teamsCheck what is said to customers.
AI for compliance teamsMarketing teamsShare customer messages and tone of voice.
AI for marketing teamsWant to see a reply checked before it is sent?
Book a demo and we will walk through a draft, a check, a sign-off and the record, using example data.
Straight answers.
Can AI answer customers safely?
Yes, when it answers from approved sources, is checked against your rules, and hands over to a person for anything sensitive. A tribunal has held a company to what its chatbot said, so the company remains responsible for every reply.
Will customers accept AI in customer service?
Most will if they can reach a person. Gartner found 87% of customers say a company using generative AI for service must provide access to a human agent.
What does GenWA do for customer service teams?
GenWA Comms drafts replies and checks them against your rules, Contact helps advisers during live calls, Guard reviews customer communications, and the Log records every reply and exception.
Does AI send replies without a person?
Only where you decide. By default a trained person approves what is sent. Some low-risk reply types may go out after an automatic check if you choose, and that choice is recorded.
Can GenWA help with complaints?
Yes. It records what was said and why, and routes complaints to a named person. Skill Teams for complaints are being built. People decide complaint outcomes.
Is this suitable for a small team?
Yes. A service team in a business of ten people or more can start with one reply type, a short list of rules and a named approver.
Where is the record kept?
It is held in your AWS or Azure account, or hosted by GenWA on AWS as agreed. We agree which in writing before we start.
Faster replies. Checked before they go. A person within reach.
Reply checks cover AI used through GenWA. People decide on credit, hiring and medical matters.