01 | Turn procurement into one real job
‘We need an AI tool’ is not a buying problem. ‘Turn three customer interviews into a source-linked requirements brief before Friday review’ is. The latter defines inputs, output, frequency, owner, and completion criteria.
Record current time, tools, and common rework. New software creates verifiable value only when it lowers cost or improves quality on the same job.
- How often does the job occur each month?
- What output is acceptable, and which errors are unacceptable?
- Who operates, reviews, and owns failure risk?
02 | Set elimination criteria before evaluating strengths
Demo surprises easily distort decisions. Write three hard constraints first: where data cannot go, which systems must integrate, and the annual cost ceiling. A candidate that fails a hard constraint does not enter scoring.
This intentionally reduces the shortlist. For a small team, three qualified candidates usually outperform ten shallow trials because context switching, setup, and migration are the expensive parts.
03 | Calculate true first-year cost
The monthly fee is only visible cost. First-year cost also includes seat growth, overage, required add-ons, migration, training, implementation, payment rates, and exit exports. Promotional pricing belongs on a separate line, not as the long-term baseline.
Model three scenarios: current team, doubled headcount, and tripled usage. If either triggers a higher tier, discover that before the trial—not at renewal.
- Regular renewal price and promotion end date
- Upgrade thresholds for seats, contacts, projects, storage, or credits
- Export, cancellation, and migration cost
04 | Run a seven-day pilot with the same job
Use the same input, owner, and scorecard for every candidate. Avoid vendor-prepared examples. Real material reveals permission, format, language, export, and collaboration issues.
Include an operator, reviewer, and administrator at least once. Operators see speed, reviewers see errors and traceability, and administrators see permissions, billing, and exit risk.
05 | Keep ‘do not buy’ as a formal option
If the pilot does not materially improve the job, keeping the current process may be best. Include ‘keep current process’ in the decision sheet to avoid buying merely because time was invested in trials.
Record four things: why the choice was made, which assumptions remain untested, when to review, and what triggers exit. Software selection is an operational decision that should be revisited.