01 Codify the playbook
- You bring
- Your sales SOP, or a top closer’s head
- What it becomes
- A staged sales flow — every stage with its own goal, script and exit conditions
- How you know it is done
- One full customer path runs end to end
What do you need to launch an AI sales agent — and how long?
Bring one thing: the sales playbook you have already proven. The four steps after it are handled on the platform — your method becomes a staged sales flow, roles are cast to your process, the review workbench scores it pre-launch, and you draw the takeover line. No prompts, no flowcharts. It all runs in a sandbox until go-live, nowhere near a customer.
You bring four things
None of it is new. It already exists in your business, just scattered across people.
Your playbook
How to open, how to read intent, how to answer the usual objections, when to close. A written SOP is ideal; your best closer’s head works too.
Roles and limits
Who takes first touch, who follows up, who answers questions, which part stays with a person. Your process, not a template.
Knowledge & assets
Course schedules, service details, policy terms, how you quote. Each bound to the role that needs it — and only that role.
Standards & red lines
What counts as a good conversation, and what must never be said. You need both: together they become the test cases.
Configured on the platform — no prompts, no flowcharts. Connection depends on where customers reach you — a site chat widget, WeCom or connected social messaging; it also connects to your existing CRM.
Four steps, from your playbook to live
The four stages from the home page, as they land on your side: what you bring, what it becomes, when it is done.
02 Scale the execution
- You bring
- Roles, and where each one stops
- What it becomes
- A team split by role — a lead agent routes; booking, follow-up and Q&A each hold their own
- How you know it is done
- Every role knows what it takes and where it stops
03 Compound the learning
- You bring
- Your standards and a batch of problem cases
- What it becomes
- A set of test cases — hard rules judged by machine, quality judged by a model
- How you know it is done
- No pass on the review workbench, no launch
04 Humans where they matter
- You bring
- The takeover line: deal size, customer type, risk
- What it becomes
- A per-customer takeover switch, handoff backstage
- How you know it is done
- When it fires, the customer notices nothing
Speed comes down to three things: is the playbook written down, how many roles, is the material there.
Start with one slice, not the whole funnel
The takeover switch is set per customer, not globally — so a pilot can be small. As small as one batch of new leads.
High volume, repetitive
The slice you run dozens of times a day: qualifying, booking, the usual questions. The more repetitive, the sooner it goes.
A clear finish line
A booked visit, a trial class, a held slot. With a finish line you can tell if it worked.
Failure you can afford
New leads and small enquiries go to AI first. Big deals and old customers stay with people. You draw the line.
Each of the six industries has its own first slice — see the industry pages.
Three gates, if it gets it wrong
One before launch, one during, one after.
Before: it sits an exam
Hard rules are judged by machine: what it must say, what it must never touch, which tools it had to call. How well it talks is scored by a judge model.
During: a human takes it back
Conversations go back to a person one customer at a time, backstage — to the customer it is the same person.
After: every turn replays
Why it answered that way, which tool it called, which knowledge it hit, how long it took. Turn by turn.
Save a new version, apply it once it passes review — the new version is applied explicitly, and you can roll back.
Three reasons to wait
Three reasons to wait
- You want answers, full stop
- Few deals, all relationships
- No playbook that works yet
Not every company should start now. In three cases we will tell you to wait.
- We do Q&A — it is one of the four roles. But if answering the question ends the job, with no intent to read and no follow-up, there are cheaper options.
- Very high ticket, a few dozen deals a year, closed over dinner. Customisation has nothing there to pay it back.
- If the method still changes and everyone tells it differently, prove it out first. Nothing to build from.
Rather build it yourself, or still comparing vendors? These two pages fit better.
Five things to test in the demo
Forward this page to whoever runs the evaluation. A demo is short — do not spend it on a feature tour.
Play a customer and talk to it
Left is the chat; on the right, role handoffs in real time. The trial runs in a sandbox, not real customers.
Pick your hardest objection
Does it answer with your method, or with something generic that sounds right?
Ask for the scorecard
Which hard rules passed, which failed, and what the judge model said. The failures are the interesting part.
Ask for a full replay
Take a single turn and ask: why that answer, which tool, which piece of knowledge, how long.
Draw a takeover line
Name a deal size or a customer type, then watch the customer’s side when it fires.
One more worth asking: which metrics to watch. Two we will answer now: no on-premise deployment — it is multi-tenant SaaS; and it connects to your existing CRM.
Bring your playbook — watch it become a team
A demo takes about 30 minutes and covers three things:
- 01Map your current sales playbook and find what AI should take
- 02Show how your method becomes an AI closing team
- 03Run both ledgers — conversion and cost — on your own numbers
Or contact us directly business@toppp.ai