A new tool is worth considering when it addresses a problem you can describe. Start with a task that takes too long, information that gets lost or a repeated mistake. Then compare the proposed change with what you do today.
Start with the incident, not the category
“We need AI” is not a useful brief. Neither is “we need a CRM” or “we need to automate.”
Start with something that actually happened:
- Two enquiries went unanswered because they landed in different inboxes.
- The same booking details were copied into three systems.
- A customer could not tell whether the information on the website was current.
- The weekly report took half a day and still contained mistakes.
Choose a recurring problem from the last month. Record where it happened, how often it happens, what you do today and what it costs in time or rework. That gives a trial something to measure.
Ask four questions
Before adopting anything new, ask four questions.
1. Does it solve the named problem?
Run the proposed tool on real examples, including awkward ones. If a demonstration needs perfect inputs and an expert standing beside it, it is not yet a business process.
2. What new failure does it create?
Automation can make the same mistake faster. AI can produce a confident answer without good evidence. A new platform can become another inbox nobody checks.
List what happens when the tool is wrong, unavailable or abandoned. Decide which actions still need a person to review them.
3. Can the result be checked?
“It feels more efficient” is hard to manage. Time spent, unanswered enquiries, error rate, completed follow-ups and customer wait time can be checked.
Choose one measurement before the trial begins. Keep the old process available until the new one has earned trust.
4. Who will maintain it?
Somebody must maintain access, update the instructions, notice failures and decide when the system should stop. If the answer is “the tool,” the work has not been designed yet.
For AI projects, the NIST AI Risk Management Framework Playbook offers guidance under four functions: Govern, Map, Measure and Manage. It is voluntary and can be used selectively. The questions above are a practical starting point for a small trial, not a claim of compliance with the framework.
A short trial should be allowed to fail
Keep the first trial small enough to review and reverse.
Choose one bounded task. Use non-sensitive or properly minimised data. Keep a human confirmation before messages, payments, publishing or account changes. Record the failures as carefully as the successes. Decide the stop condition before starting.
If the trial fails, stop it and record why.
If it works, test the next task separately. A useful result on one task does not establish that the same tool will work across the business.
Decide what happens after the trial
Keep a short record of the original problem, the result and the remaining limitations. Include ongoing fees, maintenance time and the way to export your information if you stop using the tool.
A clearer form or a change to an existing system may be enough. A new tool should earn its place through a result you can check. Our lead management and CRM service starts with the enquiry process you need to improve.
Related: CJAX Labs services and how we work.