Quick Answer

The biggest reasons admissions AI deployments underperform aren't technical — they're decisions that weren't made clearly before launch: which inquiry sources fire the agent, what counts as a qualified lead, how escalation works, who owns the knowledge base. This checklist covers the 10 decisions every team should make before go-live.

Deploying an admissions AI agent is not primarily a technical project. The platform matters, but the decisions you make during setup have more impact on conversion outcomes than the platform you choose. The institutions that see strong results from day one are the ones that work through these decisions before the agent goes live, not after.

Work through this checklist before launch. Each item is a decision, not a configuration detail. Someone on your team needs to own each answer.

The 10 Pre-Launch Decisions

1. Which inquiry sources will trigger the agent, and are all of them connected?

The most common setup error: connecting only the primary web form and leaving significant inquiry volume untriggered. Audit every way a prospective student can express interest — web forms, WhatsApp buttons on your website, click-to-WhatsApp ads, open day registration forms, QR codes on printed materials, social media bio links. Every source should fire the agent within 60 seconds. Decide which sources are in scope for launch and which will be added later. Write it down.

2. What is the target response time, and who is accountable if it's missed?

The answer should be: under 60 seconds for every connected source, always. Firms contacting leads within 5 minutes are 100x more likely to make contact than those waiting 30 minutes. Agree on this target before launch and confirm technically that it's achievable across every connected source.

3. What does the agent know about your programmes, and how will you keep it current?

Build a knowledge base before launch covering: all active programmes (duration, format, start dates, fees), entry requirements per programme including non-standard pathways, funding options, the most common objections and your considered responses, and escalation criteria. Decide who owns updates to this knowledge base after go-live. Stale information is worse than no information.

4. What counts as a qualified lead?

Define this explicitly before the agent is configured. Most institutions need: programme interest, timeline (this cycle, next year, exploring), current situation (working professional, recent graduate, international), and primary concern or question. Match these criteria to what your advisors actually need to start a productive call, not a generic B2B qualification framework.

5. How will call booking work, and how much friction is acceptable?

Decide whether call booking happens inside the WhatsApp conversation or via an external link. The former reduces drop-off. Also decide: how many steps does the booking take? What confirmation message does the prospective student receive? When do reminders fire? Agree on the maximum acceptable friction before configuring the flow.

6. Which CRM does the agent sync to, and what data gets written?

Confirm the CRM integration before launch. Decide exactly what gets written to the contact record: conversation transcript, qualification data, lead score, programme of interest, scheduled call details. Define the CRM field mapping. An agent that qualifies leads well but doesn't write to the CRM creates as much work as it saves.

7. What are the escalation triggers, and what happens when one fires?

Define the specific signals that cause the agent to hand off to a human: mentions of financial hardship or fee disputes, visa and immigration questions, accreditation concerns, expressions of frustration, explicit requests to speak with a person. Decide how escalation works technically — real-time handoff, a flagged conversation, an email alert to the advisor — and what the prospective student is told when it happens.

8. What does the cold lead re-engagement sequence look like?

Decide: how many days after last contact before the first re-engagement message? What triggers a second attempt? What events should trigger a re-engagement regardless of time (new intake, deadline approaching, open day)? What's the maximum number of automated re-engagement attempts before the contact is passed to manual follow-up? Don't leave this to be figured out post-launch.

9. How will no-shows be handled?

Booked calls that don't happen need a recovery flow. Decide: how long after the missed call does the agent send a re-engagement message? What does that message say? How many automated attempts before the advisor steps in? No-show recovery is one of the highest-ROI automations available — configure it before launch.

10. What metrics will you track, how often, and who reviews them?

Agree on four core metrics before launch: response rate, qualification rate, call booking rate, and human handoff rate. Decide who reviews them, on what cadence (weekly for the first 8 weeks, then monthly), and what threshold would trigger a configuration review. Without this agreement, the agent becomes a black box.

Frequently Asked Questions

What's the most important item on this list?

Item 1 — inquiry source coverage — has the highest impact on conversion volume. An agent with a perfect knowledge base that's only connected to one of your five inquiry sources is leaving the majority of its potential value uncaptured. Start the audit early.

How long should setup take?

For institutions with a clear programme structure and an existing CRM, 3–5 weeks from scoping to go-live is typical. Larger institutions with many programmes and complex entry pathways may need 5–8 weeks. The knowledge base build and testing phase is usually where time is underestimated.

Should we do a soft launch before connecting all sources?

Yes. Run with one or two inquiry sources for the first 1–2 weeks, review conversation logs, fix gaps, then connect the remaining sources. This prevents knowledge base issues from affecting your full lead volume before they've been caught.

Who should own item 3 (the knowledge base) on an ongoing basis?

Typically a programme coordinator or admissions manager — someone with real-time visibility of programme changes, fee updates, and intake date changes. The knowledge base should be treated like a living document, not a one-time setup task.

What if we haven't resolved all 10 items before a planned launch date?

Prioritise items 1–5. A launch with a strong first five decisions and incomplete items 6–10 will still outperform a delayed launch. Build a post-launch roadmap for the remaining items and a clear timeline for completing them.

Getting these decisions right before launch is the difference between a deployment that improves every week and one that creates as many problems as it solves. Most of them take a meeting, not a month.