OXHN Partners / Practical AI
Decide whether an AI workflow is worth testing
An AI opportunity sprint helps turn a recurring operational problem into a decision: improve the process, configure an existing tool or test a focused AI workflow. OXHN’s existing sprint scope produces a workflow map, an options comparison and a written next-step recommendation—even when the recommendation is not to build.
When is a sprint useful?
Use this starting point when the bottleneck is recognizable but the right intervention is not. The process owner should be able to describe the current steps, review effort, handoffs and constraints. The purpose is to clarify what needs to be learned before implementation, not to label every task an AI opportunity.
What should the assessment examine?
- Map where inputs come from, who acts on them and who approves the result.
- Separate missing information, waiting and ownership problems from work a tool might help with.
- Compare process improvement, existing tools and a bounded pilot against the same decision criteria.
- Define what better would mean and what evidence could justify testing, deferring or stopping.
What do you leave with?
- A workflow map that makes the current task and handoffs visible.
- An options comparison grounded in operating constraints.
- A written next-step recommendation, including the possibility of no build.
The sprint clarifies a decision before implementation. Any subsequent pilot is scoped separately, based on what needs to be learned.
A synthetic assessment decision example
Manually constructed teaching example; synthetic and unmeasured. Not client work or model results. No model was run and no performance was measured.
- Task boundary and workflow map
- Consider preparing an internal handoff checklist from an approved project brief—not drafting professional advice or approving delivery. Map: receive brief → check required inputs → draft checklist → owner reviews → hand off. Missing inputs return to the source owner.
- Unknown baseline
- Current preparation time, correction effort, missing-input frequency and waiting time are unknown. A faster or cheaper process cannot yet be claimed.
- Options comparison
- A manual checklist may expose missing inputs without new software. An existing tool may help if it fits permitted data handling. An adapted or custom workflow would need evidence that its extra review, integration and maintenance are justified.
- Evidence to gather
- The process owner identifies permitted examples and source versions. A reviewer records required checklist items, omissions, corrections, active effort and waiting across ordinary and difficult cases. Check whether missing information, rather than drafting, is the bottleneck.
- Illustrative written recommendation
- Defer an AI build. First clarify required inputs and ownership with a manual checklist and establish a baseline. Reconsider a bounded pilot only if evidence shows a remaining task worth testing and a reviewer can check it. This is a constructed decision, not an observed outcome.
Use the blank scope-and-acceptance record to record your own unknowns and next decision.
How can you prepare?
Bring the process owner and a recent example you can discuss. Use the workflow-fit worksheet to note what is known and what is still an assumption. Do not enter confidential material into the booking form on Cal.com; discussion of access belongs in scoping.
What counts as evidence?
Describe the existing process before estimating improvement. Review effort, error handling and waiting matter alongside generation time. The build-or-buy guide provides comparison records; its illustrative economics are hypothetical, not measured savings or OXHN prices.
Questions before proceeding
Does the sprint guarantee an AI implementation?
No. A simpler process or an existing tool can be the right recommendation. The decision comes before a commitment to build.
What happens if a pilot is justified?
Agree its scope, permitted inputs and evaluation criteria before building. The focused workflow pilot explains how to compare quality, effort and failure cases.
How long does it take and what does it cost?
An opportunity sprint scopes the decision before implementation. Effort depends on how clearly the workflow is documented, the variety of inputs and handoffs, and how many plausible alternatives need comparison. Access to permitted examples and the process owner’s and reviewer’s availability affect elapsed timing as well as active work.
Before starting, agree the workflow map, options comparison and written recommendation to be delivered; preparation responsibilities; permitted inputs; assumptions and exclusions; decision owner; timing and fee basis. Bring a non-sensitive workflow description, known constraints and the people who can explain and review the work. Building integrations, running a pilot and ongoing operation are separate scopes—not automatic next phases.
What happens after the first conversation?
First, decide whether there is a fit: an opportunity sprint, a focused pilot, a simpler process change or no engagement. Booking a conversation does not commit you to a pilot or wider rollout.
If there is a suitable next step, agree a written scope before work begins: the included task and exclusions, permitted inputs and access, outputs, evaluation or decision criteria, reviewer and workflow owner, handoff, timing and fees. Changes to the task, integrations or evaluation are assessed against that scope, not automatically included. Ongoing operation and support are separate responsibilities to agree.
Discuss the decision
Describe one recurring task and what you need to know before changing it. See the AI consulting approach and OXHN’s approach.
Book a 30-minute conversationOpens Cal.com scheduling. Please don’t include confidential information.
Prepared by OXHN Partners · Materially updated .