An opinionated readiness audit built by an operator. Twenty questions across five dimensions, a weighted score, and a personalized 90-day prep plan. Two questions are deployment blockers. Fail either and the audit calls it out plainly.
No email required to see your score. Email only required for the full personalized prep plan.
Question 1 of 20
Five dimensions, four-to-five questions each, mostly yes/partial/no. Two questions are explicit blockers. Fail either and the audit calls it out plainly. Takes about 4 minutes.
Score 0–100 with traffic-light status (Green / Yellow / Red), a per-dimension breakdown showing where points came from, and your top three prep priorities. All visible without an email.
Optional email gate for the full plan: every triggered task, sequenced by phase, with named owners and effort estimates. Formatted to forward to your CX lead, or to use as a leadership-review agenda.
The vendor demo always works. The pilot always looks good. The trouble starts in month four, when the easy intents have been won and the AI has to handle the edge cases: the gift recipient who can’t access the order, the chargeback that’s actually a fraud claim, the customer who threatens legal action over a $30 refund.
These are policy moments, not knowledge-base moments. AI handles them as well as the underlying decision system you’ve built: the documented escalation criteria, the refund matrix, the always-escalate list. Without those, AI either hallucinates a policy or escalates everything, which kills containment and CSAT in the same quarter.
That’s why this assessment weights SOP and policy design at 25%, the same as the knowledge base. Most readiness checklists fixate on whether you have an FAQ and an API. Those matter, but they’re not what breaks programs in month four.
Most things on this audit can score zero and still be fixable inside a 90-day prep window: missing KB articles, undocumented SOPs, untracked CSAT. Tedious to fix, but fixable.
Two things can’t. Helpdesk API access determines whether AI can integrate with your support stack at all. If your current plan tier doesn’t include the API, AI literally has no way to read tickets, update statuses, or route conversations. The integration path doesn’t exist.
Customer identity is the other one. If you can’t reliably answer “is this the same person who emailed last week,” AI can’t personalize, can’t detect repeat contacts, can’t link tickets to orders. Fragmented identity isn’t a polish issue. It breaks the foundational contract AI needs.
Both are solvable. Both take real time to solve (weeks to months). And both must be solved before deployment, not in parallel with it. Calling them out as binary blockers, rather than averaging them into a score, keeps the verdict honest.
A score is a number. A traffic light is a color. Neither tells you what to do tomorrow morning.
The personalized prep plan does. Every triggered task ties back to a specific question you scored low on, with a named owner role, an effort estimate in person-weeks, and a phase suggestion. Phase 1 (days 0–30) covers the foundational work. Phase 2 (days 31–60) covers the build. Phase 3 (days 61–90) covers enablement and dry-run.
A typical Yellow-status brand triggers around 12–15 tasks totaling 15–20 person-weeks of effort. That’s 90 days for two or three people working part-time on the prep, which is exactly the right size to scope as a Tiny Swell readiness engagement, or to run internally if you have the bench.
The AI Customer Service ROI Calculator estimates annual savings, payback period, and cost-per-ticket reduction across three scenarios. Five inputs, CFO-ready output.
Run the ROI calculatorWhat people ask before running the assessment, answered so you can decide whether it's worth your 4 minutes.