A working AI workflow and a scale decision in 14 days
We build and measure a working workflow on your real data before you commit to a larger deployment.
What you provide
- One workflow
- One project, plant, site, or operational unit
- A representative document set
- 20–30 real questions, tasks, or historical examples
- One workflow owner
- One technical or security reviewer
What you receive
1. Workflow map
2. Manual baseline
3. Working AI workflow
4. Citation-linked outputs
5. Evaluation scorecard
6. Human approval and escalation rules
7. Failure-case log
8. Before/after measurement
9. Scale recommendation
Good fit when
- The workflow happens repeatedly
- Documents are the main source of delay
- You can share representative data
- A human currently reviews the work
- The result can be measured in two weeks
- There is a path to a larger deployment
Not a fit when
- You want a generic AI strategy
- No data can be shared
- No one owns the workflow
- You expect a free enterprise-wide proof of concept
- Full ERP, SCADA, BIM, or OT integration is required before value can be tested
- The output has safety or regulatory consequences but no human approval process exists
Acceptance criteria (examples, not guarantees)
- 20–30 representative tasks evaluated
- Citation correctness measured
- Time-to-completion compared with the current process
- Unsupported answers and failure cases documented
- Human reviewer approval rate measured
- Security and deployment constraints recorded
- Scale recommendation based on evidence
The sprint establishes the customer-specific result. Acceptance criteria below are examples of what we measure, not a guaranteed percentage.
Commercial
Paid, fixed-scope engagement. Fixed fee, agreed before kickoff and confirmed in the assessment call. The fee is scoped to document volume, deployment boundary, security requirements, and workflow complexity.
If the 14-day proof works, we expand it into a supervised 30–90-day deployment. Deployment from US$20K.
FAQ
What does a 14-day AI proof sprint include?
A paid, fixed-scope engagement that maps one document-heavy workflow, builds it on your real data, returns cited outputs, records human approval rules, and ends with a measured scale recommendation. You provide one workflow, a representative document set, 20–30 real tasks, and named reviewers.
How much does the 14-day proof sprint cost?
A fixed fee, agreed before kickoff and confirmed in the assessment call. It is scoped to document volume, deployment boundary, security requirements, and workflow complexity. If the proof works, deployment starts at US$20K.
When is the proof sprint not a fit?
Do not start here for a generic AI strategy, when no data can be shared, when no one owns the workflow, or when full ERP, SCADA, BIM, or OT integration is required before value can be tested. Safety and regulatory outputs still need a human approval process.