AI Workflow Audit
Fixed fee · 1–2 weeks
We map one team's workflows, identify 3–5 automation candidates, and deliver a scored roadmap with ROI estimates. Yours to keep, whoever builds it.
From $2,500
Book an auditAI Agents & Automation
AI agents built for production. And kept there.
Most agents die somewhere between the demo and the day-to-day. Ours are built with guardrails, evals and monitoring from day one, and we stay on to run them.
The demo is not the product
Wiring an LLM to a workflow and getting an impressive demo takes a weekend. Then reality arrives: edge cases the prompt never met, a model update that quietly changes behaviour, an integration that times out at month-end volume, and costs nobody is watching.
A production agent is an operations problem. It needs error handling, evals that catch regressions before your customers do, audit logs, cost ceilings, and someone on call. We have been doing that kind of work since 2003. The AI part is new. The discipline isn't.
Pick your starting point
Pilot Agent
Fixed scope · 4–6 weeks
One workflow, one agent, human in the loop. Measured against a baseline so you know whether it actually works.
From $10,000
Scope a pilotProduction Build-out
Project pricing
The pilot, hardened: evals, guardrails, integrations, dashboards, security review, and deployment into your infrastructure.
Scoped per project
Discuss a buildManaged AgentOps
Monthly retainer
Monitoring, eval regression on every model update, cost control, and incident response. Agents don't stay working on their own. We keep them working.
From $1,500 / mo
Ask about AgentOpsHow an engagement runs
-
01
Audit
We sit with the people doing the work, map the workflow end to end, and score every automation candidate by ROI and risk.
-
02
Pilot
We build one agent for the highest-scoring workflow, human in the loop from day one, and measure it against the baseline recorded in the audit.
-
03
Production
The pilot gets hardened: evals, guardrails, integrations, dashboards, and deployment into your infrastructure or onto ours.
-
04
Managed
We monitor runs, re-run evals when models update, watch the costs, and answer the pager when something drifts.
The track record
We've shipped agent systems for clients (see the Media Regent AI agents project) and we use the same tooling in our own work every day. The numbers behind the offer:
- 20+ Years in production
- 150+ Projects delivered
- 40+ Clients worldwide
Common questions
How much does a custom AI agent cost?
A fixed-fee AI Workflow Audit starts at $2,500. A Pilot Agent, one workflow built and measured, starts at $10,000. Production build-outs are scoped per project, and Managed AgentOps retainers start at $1,500 per month. You know the number before we start.
How long does it take to get an agent into production?
An audit takes one to two weeks. A pilot agent takes four to six weeks from kickoff to measured results against a baseline. Hardening beyond the pilot depends on your integrations, but you will have a working, supervised agent by the end of the pilot, not a slide deck.
Do we need our data cleaned up first?
No. The audit works with your systems as they are, including the messy inboxes, spreadsheets, and legacy databases. Part of its job is finding the workflows where today's data is already good enough, and flagging where cleanup would actually pay off.
Will a human stay in the loop?
Yes, by default. Every pilot ships human-in-the-loop: the agent drafts, a person approves. Autonomy grows only where eval results justify it, and everything is logged, so your compliance team can see exactly what the agent did and why.
Can agents run on our own servers or with private models?
Yes. Private, on-premise LLM deployments are an existing iinix service. If your data cannot leave your infrastructure, we deploy open-weight models inside it and build the agent around them.
Got a workflow that eats hours?
A fixed-fee audit is the cheapest way to find out what an agent could do about it, and whether it's worth doing at all.
Start with an audit