Define the exception
Pick one approval bottleneck, status discrepancy or multi-step administrative handoff. Agree on expected outcomes.
Bolt · Exploratory pilots
One workflow.
Under control.
A tightly scoped feasibility study for AI-assisted workflows with explicit review boundaries, exception handling and traceable decisions. Bolt is an early macOS prototype, currently tested internally. Independent customer validation remains to be done.
Email only · Fully remote · No telephone calls, video meetings or in-person sessions
Narrow scope
Prove one useful thing first.
We seek an operational owner who can describe a recurring, non-critical exception in writing. We start with fictional test inputs and define a measurable acceptance criterion before discussing real systems.
Bound the actions
Specify what may be suggested, what must await review, and what is denied. Never assume unrestricted execution.
Inspect the evidence
Measure pass/fail on synthetic cases, review exceptions and document limitations. No customer ROI is presumed.
Written-first collaboration
No meetings required.
01
Describe by email
State the problem without confidential records, customer information or access credentials.
02
Agree boundaries
Review a proposed scope, written milestones and an indicative non-binding budget.
03
Test synthetic inputs
Subject to separate legal and written agreement, run a constrained feasibility exercise.
04
Receive findings
Get a written outcome summary, failure cases and any prerequisites for future integration.
Current technical evidence
What we can substantiate.
63/63Local Python tests passed in a separate experimental evidence agent. The local Qwen model returned three citation-validated suggestions for fictional exception cases. Inspect the synthetic AI evidence → None of these tests represents independent commercial validation.
- Internal only. The earlier Swift Bolt test harness used 14 fictional logistics events; five entered scripted approval/rejection checks. A separate Python experiment now calls a real local Qwen model on four other fictional orders. No real employee approved an external operation.
- Not production. No WMS, ERP, industrial controller, carrier or customer application is integrated.
- Validation outstanding. Developer tests passed, but independent user testing, security certification and commercial benefits are unproven.
- Separate public example. MIT demonstration of action decisions is not the proprietary macOS Bolt product.
Indicative scope
Small and measurable.
€3,000–€5,000
Illustrative, non-binding range for a 2–3 week feasibility exercise. Final scope, timeline, taxes and any price depend on technical fit, resources, written agreement and legal readiness. This page is not a fixed-price purchase offer.
€6,500
Illustrative, non-binding expanded feasibility package: written requirements mapping, one synthetic demonstration workflow, predefined acceptance tests and a findings report over an estimated 2–3 weeks. This does not include production integrations, independent security certification or guaranteed savings. The final scope, timing, taxes and price require feasibility review, legal readiness and separate written agreement.
For US buyers, we also have a separate illustrative USD $6,500 / 10-working-day synthetic-only feasibility scope. These are different exploratory packages, not a currency conversion or a fixed price guarantee.
Unfire is a founder-led technology initiative in Spain, not a separately incorporated company. No remunerated work would begin before resolving relevant Spanish tax, Social Security and benefits obligations.
Email enquiries
Start with the problem.
Tell us which process is failing, how a reviewer would know it was resolved and whether a written exploratory project could help. Please do not send passwords, private documents, production credentials or personal data.
hello@unfire.technology ↗Communication exclusively by email. No calls, video conferences, visits, compulsory live demonstrations or unapproved access. No support-time guarantee or production SLA is implied.