OPERATING NOTES · 20 SEPTEMBER 2026
How to assess a technology capability at its current stage
Define the claim, examine the evidence and choose a proportionate next step.
A technology discussion becomes more useful when everyone agrees what is being assessed. A research result, a prototype and a service used in routine work answer different questions. Before discussing scale, define the capability, the setting in which it has been demonstrated and the evidence still missing.
Start with one specific claim
Replace “the platform automates the process” with a bounded statement: which task, which inputs, which users and which conditions? Ask what happens outside that boundary. A demonstration can be useful evidence for a narrow claim without supporting a broader deployment decision.
Research: what has been learned?
Ask for the question investigated, the method used and the limits of the result. Which conditions mattered? What remains uncertain? A useful next milestone is a test that could change the team's understanding. A research result can justify further investigation without supporting a delivery date.
Development: what can be repeated?
For a prototype, ask a second person to follow the documented process on an agreed case. Record required setup, manual intervention and failures. Define what would count as a successful repeat before the demonstration. This is a practical exercise, not a universal readiness score.
Operations: who keeps the work running?
For a service intended for everyday use, establish responsibility for support, incidents, changes and recovery. Ask how a failed task is noticed and corrected, what information is retained and which commitments apply today. Keep proposed support arrangements separate from those already available.
Make a one-page capability record
- The exact capability claim.
- The environment in which it was assessed.
- The evidence and its date.
- Known limitations and unavailable evidence.
- The accountable owner.
- The next decision and its conditions.
Leave unavailable evidence marked unknown. Update the record when the underlying evidence changes, rather than refreshing its date for appearance.
A partner can then choose a proportionate next step: further research, a bounded development exercise or an operational evaluation. Different capabilities within the same organisation may require different decisions. The objective is an informed commitment with explicit conditions, not a single label for an entire business.
For the principles behind this exercise, see Oxford Intelligence’s operating principles. For a related planning tool, use the applied AI pilot evidence checklist. To discuss a possible technology partnership, visit our partnership section.
This is editorial guidance, not a certification standard or an assessment of a named venture. Oxford Intelligence Limited is an independent company and is not affiliated with or endorsed by the University of Oxford.