烈火 · The offer

The Airekka AI audit

A structured diagnostic of how work actually moves through your company — people, systems, documents, and handoffs — so you can decide where AI belongs before you spend on a build.

烈火 · What it is — and is not

An operational audit. Not a model-ethics review. Not a demo.

When people hear “AI audit,” they often picture one of two things: a compliance review of a production model, or a vendor walking through ChatGPT use cases. This is neither.

We sit with the people who do the work. We watch the handoffs. We look at the ERP, EHR, CRM, and the unofficial spreadsheets that fill the gaps. Then we write down where AI would pay off, where the process has to be fixed first, and where it should stay human.

That is the same discipline as the engagement we are in now: support already lived in client-facing mailboxes, a third-party ticketing tool was the extra hop, and the recommendation was not “add AI on top of the ticket queue.” It was start at the mailbox, close what AI can close on a private, in-house system, and escalate the rest. The audit is that habit, named as the front door.

烈火 · How we run it

Four moves. One written brief.

  1. 01

    Intake

    You name the process that hurts — or we help you name it. One painful operation is enough to start. A wider map is available when the first slice proves useful.

  2. 02

    Map

    We interview operators, watch the work, and draw the real path — including the unofficial steps. Systems, documents, and exception handling all count.

  3. 03

    Score

    Each candidate is scored for volume, exception rate, data readiness, integration cost, irreversible risk, and payoff. Pretty demos do not get extra points.

  4. 04

    Recommend

    You get a go, fix-first, or don’t-automate call on each item, with a sequence for the next 90 days. If a first pilot is justified, the brief includes a scoped recommendation for that too.

烈火 · How we actually work

This is systems analysis. The industry just keeps renaming it.

There's a lot of mystique around AI assessments — frameworks, certifications, colored belts. Underneath the branding, the real work is older and simpler: sit with the people who do the job, map how the work actually moves, find where it breaks, and judge honestly what's worth changing.

That discipline is called systems analysis, and it has underpinned every serious systems project for forty years. We run the fundamentals. Where a standard name helps us talk to your team — or where a regulator needs something they recognize — we use it. But you're not paying for an acronym. You're paying for someone to look clearly at your operation and tell you the truth about it.

Here's what that looks like, chapter by chapter.

  1. 01

    Workflows — the real path of work

    We map how work actually moves, including the unofficial steps and the SOP nobody follows. We look for where time dies, where exceptions pile up, and where two teams argue about whose number is right.

    The fundamentals: current-state process analysis. You may hear it called value stream mapping or process discovery.

  2. 02

    Systems — what you already run

    We inventory what’s running and, more importantly, how it connects — where systems talk to each other cleanly, and where a person is the integration, quietly re-keying data between them.

    The fundamentals: systems and integration analysis. Sometimes branded as portfolio assessment or integration mapping.

  3. 03

    Documents — what’s trapped in files

    We trace the unstructured stuff — PDFs, scans, emails, portals, lot records — into the actions it feeds. If people are reading so a system can act, that’s where we look.

    The fundamentals: data and information flow analysis.

  4. 04

    People — who handles the messy cases

    We talk to the people who do the work and find out what they actually decide, especially on the exceptions. Some of that judgment can be supported. Some of it must stay human. Knowing the difference is most of the job.

    The fundamentals: requirements elicitation and exception analysis — the oldest technique there is, sometimes dressed up as stakeholder discovery or working sessions.

  5. 05

    Risk — whether it can be trusted here

    We assess data quality, integration cost, irreversible actions, and compliance exposure. Not whether a demo looks good — whether this can be trusted in your operation, with your risk tolerance.

    The fundamentals: feasibility and risk assessment. For regulated work we align this to the NIST AI Risk Management Framework, so your compliance team has something they recognize.

  6. 06

    Sequence — what to do first

    We turn findings into an order: what to do first, what to fix before AI touches it, and what to leave alone. Each candidate gets a go, fix-first, or don’t-automate call, with the reasoning.

    The fundamentals: gap analysis and prioritized roadmapping. This is the analyze-and-decide work; building, if it ever happens, is a separate engagement.

What you walk away with

A written brief you own — the current-state map, the findings, the scored recommendations, and a 90-day sequence. Plain language your team can act on, whether that's with us, on your own, or with someone else entirely.

The audit is the audit. If a recommendation earns a go and you want it built, that's a separate conversation with its own scope and its own quote — from us, or from someone we point you to. We don't fold the build into the audit, and we don't need you to buy one to make the other worth it.

烈火 · The recommendation

Every candidate gets one of three calls.

Go

The process is stable enough, the data is reachable, and the payoff is real. This is a candidate for a scoped pilot.

Fix first

AI would amplify a mess. Clean the handoff, the master data, or the exception path — then re-score.

Don’t automate

Low volume, high judgment, thin data, or unacceptable risk. Leave it with people. That is a finding, not a failure.

You own the brief. Run it internally, hire someone else, or ask us to pilot the first item that scored a go. We do not treat the audit as a trap to force a build.

烈火 · Read first

The AI Audit Guide

What an Airekka audit actually covers, how we score a candidate, and what you walk away with — before you ever talk to us.

Download the guide
烈火 · The worksheet

Not ready for a call? Score it yourself first.

The AI Fit Worksheet is the same questions we ask on the call, in a form you can run on your own. We'll email it to you as a PDF.

We'll email you the worksheet and nothing else. See our privacy policy.

Name the process. We’ll tell you if it is a candidate.

Response within one business day. Toledo, OH and Orlando, FL.

Start with a conversation