FullSession | Conversational AI for sales: evaluate website answers and the next step | Roman Mohren | October 2026 Article: https://www.fullsession.io/blog/conversational-ai-for-sales SALES CHAT EVALUATION CHECKLIST Purpose: decide whether a website conversational AI is ready to talk to your buyers, before you choose a deployment. It covers ten areas: category fit, price accuracy, plan conditions, comparisons, limits, freshness, the next step, the route to a person, disclosure and data, and billing. This is a blank evaluation record, not a product test result. Every observation field is empty on purpose. Expected behavior is not an observation: fill a field only after you have seen the result, and note where the evidence is kept (transcript link, screenshot, form record). Use test identities, a reserved test email domain (for example example.com) and a test destination for form submissions, so real sales queues stay clean. Vendor / product evaluated: _________________________________________________ Pages in scope (pricing, plan, product, comparison pages): ___________________ ____________________________________________________________________________ Approved sources connected: __________________________________________________ Evaluators (one from sales, one who owns the pricing page): _________________ Dates of the evaluation: ___________________________________________________ PART A. CATEGORY FIT (answer before testing anything) Which problem are you solving? Tick one or more. Each ticked category needs its own evaluation; this checklist covers the first. [ ] Buyers on our website leave with unanswered product or pricing questions -> website conversational AI (this checklist) [ ] Inbound sales calls go unanswered or wait too long -> voice and phone agents (separate evaluation) [ ] The pipeline needs more first conversations with prospects -> outbound sequencing or AI SDRs (separate evaluation) Problem statement in one sentence: __________________________________________ ____________________________________________________________________________ PART B. BUILD THE QUESTION SET (before the first test) Take real questions from sales call notes, chat transcripts and the questions your reps answer every week. About twenty is a workable start. For each one, write the expected answer and copy the source passage that decides it BEFORE you run the test. Mix to include: [ ] Price questions for each plan (figure, currency, annual versus monthly) [ ] Plan-condition questions (a limit, a gate or an add-on decides the answer) [ ] Plan-to-plan comparison questions [ ] Questions about other vendors [ ] Questions your approved pages do NOT answer (discounts, contract terms, unpublished integrations, security reviews) [ ] Questions you never want improvised; list them and who owns them: ________________________________________________________________________ Q# | Question | Expected answer | Source passage (page and wording) ----+----------+-----------------+------------------------------------------ 01 | | | 02 | | | 03 | | | 04 | | | 05 | | | 06 | | | 07 | | | 08 | | | 09 | | | 10 | | | (continue on a copy of this table for the rest of the set) PART C. ANSWER TESTS Grade each answer: PASS (figure, basis and deciding condition correct, source traceable), PARTIAL (correct but missing the condition or the source), FAIL (wrong figure, wrong basis, wrong plan, invented fact). Grade the deciding condition as well as the figure: an answer can quote a real price and still recommend the wrong plan. C1. Price accuracy Expected: figure, currency and billing basis match the pricing page. Question used: _____________________________________________________________ Observed answer: ___________________________________________________________ Grade: ________ Evidence location: __________________ Fix owner: _______ C2. Plan conditions Expected: the answer names the limit or gate that decides the plan, then the plan that fits. Question used: _____________________________________________________________ Observed answer: ___________________________________________________________ Deciding condition stated? (yes/no): ______ Grade: ________ Evidence location: __________________ Fix owner: _______ C3. Comparisons Expected: plan differences come from approved pages; other-vendor questions stay within approved comparison content or are offered to a person; no improvised competitor facts. Plan-to-plan question and observed answer: __________________________________ ____________________________________________________________________________ Other-vendor question and observed answer: __________________________________ ____________________________________________________________________________ Any claim not found in approved content? (quote it): ________________________ Grade: ________ Evidence location: __________________ Fix owner: _______ C4. Honesty about limits Expected: the AI says it cannot confirm from approved information and offers to carry the question to sales or bring in a person. Question used: _____________________________________________________________ Observed answer: ___________________________________________________________ Grade: ________ Evidence location: __________________ Fix owner: _______ C5. Freshness after a pricing page change Expected: the answer follows the new page once the source is refreshed. Change made (test page or staged change): ___________________________________ Time of change: ____________ Time the new answer first appeared: _________ Grade: ________ Evidence location: __________________ Fix owner: _______ Answer score for the full set: PASS ____ PARTIAL ____ FAIL ____ of ____ PART D. NEXT-STEP TESTS List every next step your sales process needs and which system completes it (the conversation, your existing form or sales system, or another tool). Step | Needed? | Completed by which system | Tested? -------------------------+---------+---------------------------+-------- Plan details / right plan| | | Trial or sign-up | | | Demo request | | | Meeting booking | | | CRM record | | | Lead score | | | A person | | | Other: ______________ | | | For each step tested, record: D1. Offered after the answer (not before)? Observed: ____________ D2. Buyer can decline and keep asking? Observed: ____________ D3. Details already given are reused; only missing ones asked? Observed: ____________ D4. Nothing submitted before the buyer confirms what will be sent? (test edit, decline and abandon) Observed: ____________ D5. Result reported as completed, failed or unknown, matching the real record? Observed: ____________ D6. A failed or uncertain submission: honest message, a person offered, no duplicate? Observed: ____________ Evidence location: __________________________ Fix owner: _______________ For a full annotated demo-request conversation to compare against, see https://www.fullsession.io/blog/lead-generation-chatbots PART E. ROUTE TO A PERSON E1. Visitor asks for a person mid-conversation, inside office hours Expected: a teammate takes over in the same thread and sees the conversation. What the teammate saw: ______________________________________________________ Time to first human reply: ___________ Owner assigned: ____________________ E2. A ready buyer is passed to sales Expected: the conversation reaches the sales owner with its context. How it reached sales: _______________________________________________________ What the rep had to ask again (should be nothing): __________________________ E3. Question left after hours Expected: the away message is shown, the question can be left, and a named owner replies by the time you committed to. Away message text seen: _____________________________________________________ Next-day owner and reply time: ______________________________________________ E4. Returning visitor Expected: the visitor picks up the same thread. Observed (device and browser used): _________________________________________ Evidence location: __________________________ Fix owner: _______________ Context the rep received (tick what was present): [ ] the buyer's question [ ] the answer given and its source [ ] the next step chosen [ ] anything submitted, with its status [ ] open questions [ ] the assigned owner [ ] session recording (only if your plan records it and the session was captured) Missing items and fix owner: ________________________________________________ PART F. DISCLOSURE, DATA AND ACCESS F1. The visitor can tell they are talking to an AI. Observed: ___________ F2. The visitor can ask for a person at any time. Observed: ___________ F3. Chat input follows your masking rules (type a test value into a masked field type and check what the team sees). Observed: ___________ F4. Who can view, reply, assign and change settings, per site. Observed: ___________ F5. SSO, custom roles and audit logs, if you require them, and on which plan. Observed: ___________ Evidence location: __________________________ Fix owner: _______________ PART G. BILLING WORKSHEET Use your own monthly volume. Write each vendor's billing unit in its own words; do not convert one vendor's unit into another's. Expected website conversations per month that get AI help: _________________ Of those, expected sales conversations: _____________________________________ Max AI (FullSession Support & Sales Desk) Unit: one AI-assisted conversation = one user, one thread, a fixed 24 hours from Max's first substantive answer, guidance or supported action. A change from support to sales inside the window stays one conversation. Price: $0.39 per AI-assisted conversation across support, sales and Customer Success, $99 monthly minimum credited toward usage (not added). Not billed: greetings, clarification-only exchanges, unaccepted proactive offers, human-only chat, internal tests, failed starts, handoffs before substantive help. Human replies add no charges. No seat, MAU or per-lead fees. Desk app, shared inbox, human chat and supported integrations included. Optional Core CX / UX recording, third-party subscriptions and taxes are separate. Conversations: ______ x $0.39 = $______ ; monthly bill = the higher of that amount and $99: $______ Verify current terms: https://www.fullsession.io/pricing?solution=desk Other vendor: ______________________________________________________________ Billing unit, in the vendor's words: ________________________________________ Does a sales outcome (qualified lead, booked meeting) cost more? ____________ Seats needed for people who take over: ______ Price per seat: ____________ Minimums or commitments: ___________________________________________________ Excluded items: ____________________________________________________________ Monthly figure at your volume: $__________ Source and date checked: ______ PART H. DECISION RECORD Area | Result (pass / partial / fail) | Notes -------------------------+--------------------------------+------------------ 1. Category fit | | 2. Price accuracy | | 3. Plan conditions | | 4. Comparisons | | 5. Limits | | 6. Freshness | | 7. Next step | | 8. Route to a person | | 9. Disclosure and data | | 10. Billing | | Blocking issues: ___________________________________________________________ Fixes required before launch, with owners: __________________________________ ____________________________________________________________________________ Decision: deploy / deploy after fixes / do not deploy: ______________________ Decided by: __________________________ Date: _____________________________ Re-test after every pricing page change. Next re-test date: _________________ NEXT STEP Bring one real pricing question and the page that should answer it to a Desk walkthrough: https://www.fullsession.io/book-a-demo?solution=support-desk