Field Discovery Sprint / one workflow / two to three weeks

One workflow. One clear decision.

We study one important workflow and tell you whether to build, buy, fix the process first, or stop. You leave with the evidence and the next-step plan.

Check your workflow See what you receive

Choosing between Replay and Discovery Sprint?

Replay starts with five completed Cases from a suitable recurring workflow and identifies a bounded next test.

Discovery Sprint is the broader option when the workflow, business case, or implementation readiness needs investigation.

Discuss Replay

FIELDRUNTIME / DISCOVERY

IS THIS FOR US?

Start with a business problem—not an AI wish list.

The Sprint is useful in three common situations.

You do not need a company-wide AI strategy. You need one workflow that matters and access to the people who know it.

The workflow is painful.

It consumes too much time, creates delays or errors, frustrates the team, or carries meaningful risk.

An AI pilot is stuck.

The demo works, but nobody is sure how to make it safe, measurable, owned, and useful in production.

A spending decision is due.

You need evidence before buying software, funding a build, renewing a product, or expanding a pilot.

WHAT HAPPENS

Three steps. No mystery.

We work with the people who own and perform the workflow. The recommendation follows the evidence—not a preferred technology.

01
SEE THE WORK
Show us how the workflow really works.
We speak with the people who do the work and map the steps, systems, handoffs, delays, exceptions, and decisions.
02
TEST THE CASE
Measure the value and the limits.
We estimate today’s cost, time, errors, and risk. We identify what software or AI could do—and what must remain human.
03
MAKE THE DECISION
Choose the next move with evidence.
We give you a clear recommendation, a practical system plan, the measures for success, and the reasons behind the decision.

WHAT YOU RECEIVE

Four useful outputs—not a strategy deck.

Everything needed to make the next decision.

The workflow map

A plain view of what happens now, who is involved, which systems are used, and where the work breaks down.

The business baseline

The current time, cost, volume, errors, rework, waiting, and risk—plus cost per accepted outcome and the numbers a future system must improve.

The control plan

Which intelligence should run locally, come from cache, or use specialist or frontier models—plus data limits, human authority, failover, and recovery.

The recommendation

A clear build, buy, fix-the-process-first, or stop decision—with the smallest sensible next step if you proceed.

INTELLIGENCE SUPPLY CHAIN

One workflow may depend on several kinds of intelligence.

Decide what to use—and what must remain under your control.

We do not assume one model should do every job. Inside the same Sprint, we determine where intelligence should come from, what it may see, how it is verified, and how the workflow continues when a supplier changes or fails.

01INVENTORY
Know every dependency.
- Map every model, data source, tool, router, outside service, and human handoff the workflow depends on.
- Define the provenance that must be recorded so every important output can be traced.
02ROUTE
Use the right intelligence for the work.
- Classify what should be local, cached, specialist, or sent to frontier intelligence.
- Calculate cost per accepted outcome—not merely cost per token or model call.
03RESILIENCE
Keep the workflow working.
- Document concentration risk across model providers, routers, and critical services.
- Define what happens when a provider changes, slows down, becomes unavailable, or fails a check.
04CONTROL
Limit exposure and authority.
- Specify what context may leave the enterprise, what must stay private, and what should not reach a model at all.
- Define what the system may recommend, do, spend, or commit—and where a person must approve.
05PROVE
Measure what actually works.
- Create proprietary evaluations based on the company’s real cases, policies, risks, and definition of acceptable work.
- Establish an outcome ledger connecting intelligence used, evidence, action, cost, approval, and business result.

The result: a provider-neutral plan that connects intelligence cost, company controls, operational evidence, and accepted business outcomes.

WHO WE NEED

A small group who knows the work.

The Sprint should not consume the organization. We need limited access to the people who own, perform, and support the workflow.

01Executive sponsor: A short opening conversation and the final decision meeting.
02Workflow owner: A few working sessions across the two to three weeks.
03Two to four operators: Sessions showing how the work is really performed, including exceptions.
04Technical or security lead: A focused review of systems, access, security, and deployment limits.

WHAT YOU CAN DECIDE

The answer does not have to be “build.”

Four honest outcomes.

Build

The workflow is valuable, specific, measurable, and worth turning into a production system.

Buy

An existing product can do the job. You leave knowing what it must support and how to evaluate it.

Fix the process first

The workflow is too unclear or fragmented to automate safely. Repair the operation before adding AI.

Stop

The value, evidence, readiness, or risk does not justify more spending. Avoiding the wrong project is a good result.

SIMPLE ANSWERS

Clear boundaries before work begins.

Common questions.

Is this a pilot?

No. The Sprint decides whether a pilot or production build is worth doing and defines the smallest sensible scope. It does not include the software build itself.

Do we need an AI strategy first?

No. You need one important workflow, a person who owns it, access to people who perform it, and a result the organization cares about.

What do we need to prepare?

Bring a few real examples of the work, any existing process documents, rough operating numbers, and the names of the people and systems involved. We will guide the rest.

Do you need production access?

Usually not. Approved examples, sample records, system documentation, and interviews are normally enough for discovery. Any access follows your security rules.

Do we have to choose one model provider?

No. The Sprint identifies what should run locally, come from cache, use specialist intelligence, or reach a frontier model. It also defines provider failover and the enterprise controls that should remain portable.

Must Field Runtime build the system afterward?

No. Your team, Field Runtime, or another delivery partner can use the decision pack. If we build it, the Sprint becomes the specification and acceptance plan.

How is the Sprint priced?

It is a fixed-fee engagement agreed before work begins. The fee depends on the workflow’s complexity, the available evidence, and the number of systems involved.

One workflow is enough to start.

Find out whether it is worth pursuing.

Tell us what the team does, where the work gets stuck, and what a better result would look like. We will tell you whether the workflow appears to fit the Sprint.

Check your workflow Not sure what to send? Start here MAE@FIELDRUNTIME.AI / PACIFIC NORTHWEST + UNITED STATES