An executive demands "100% accuracy" from a contract review assistant. What is the right response?
Why C is the answer
Turn it into measurable targets, with human review where errors are costly
Translate the goal into metrics and an operating model the business can accept. Agreeing to an impossible SLA damages trust later.
In detail
- No system can honestly promise 100 percent.
- Turn the goal into measurable targets the business can accept.
- Put human review where errors are costly.
- Agree what happens when the system is wrong.
Use case
An illustrative scenario, not a real incident.
An executive demands 100% accuracy. The architect proposes 98% agreement on a reviewed test set, with a lawyer checking high-value clauses, and the executive accepts.
See it in different ways
flowchart LR A["Business problem"] --> B["Baseline metric"] B --> C["Success target"] C --> D["Pilot"] D --> E["Decision to scale"]
sequenceDiagram participant S as Sponsor participant A as Architect S->>A: A broad ask A->>S: Problem, baseline and target A->>A: Build and measure A-->>S: Result, limits and options S->>A: Named decision
flowchart TD
Q{"How costly is an error?"}
Q -->|"Low"| A0["Automate and sample"]
Q -->|"High"| A1["Human review"]Key point: Turn wishes into targets and an operating model.
Why the other options are wrong
- A. Refuse the target outright, because LLMs are non-deterministic by design
- Refusing ends the discussion rather than solving the problem.
- B. Promise it for the pilot only, then renegotiate the target once results are in
- This delays the same impossible promise.
- D. Agree to the target, and handle any misses through support after launch
- Agreeing to an impossible target damages trust later.