AI project audit: find what works and what to fix
Already running a chatbot, automation or another AI project? I review output quality, costs and risks, then prepare a prioritised plan. If you are just getting started, I map suitable opportunities. The initial 30-minute call is free.
Audit of existing AI projects and systems
An audit doesn't have to start from scratch. If you already use AI — a chatbot, automation, AI tools — I'll independently verify whether it actually delivers what it promised. With concrete numbers.
Evaluation of a running or failed AI project: what the outcome was supposed to be, what the reality is, where they diverge and why. Cost vs. actual benefit, and a recommendation — fix, rebuild, or stop.
A review of the AI systems you use: output quality, running costs, data security and areas to assess under the EU AI Act (transparency, AI literacy). You get a clear overview of what to keep, replace or finish properly.
Mapping how your competitors publicly use AI — chatbots, automation, AI content — and comparing it with your company. You'll see where you lag, where you lead, and what's worth catching up on first.
How I assess a running AI project
We first define the original goal and the baseline without AI. We test output correctness, errors and handoffs on a sample of real cases. We compare the benefit with operating, review and repair costs. The result is a prioritised decision on what to keep, change or stop, with its assumptions stated.
Five questions you can ask now
- Which specific task was the project meant to improve, and what did that task cost without AI?
- Do you have a representative set of real cases, including difficult exceptions, with human-verified correct answers?
- How many cases does AI resolve correctly without correction, hand over to a person or get wrong?
- What is the full cost per correctly resolved case, including operation, review and corrections?
- Who controls data access, monitors errors and decides when to stop or change the system?
Acceptable error and cost thresholds depend on the use case and should be set before testing. For a broader framework, see the NIST AI Risk Management Framework.
Starting with AI: what the assessment includes
First I understand your business, then I design AI solutions. Not the other way round. The audit reveals where AI brings the most value — and where it doesn't.
We walk through your key processes step by step. I identify repetitive tasks, bottlenecks and the areas with the biggest AI potential.
For each process I assess whether AI makes sense. Some tasks AI handles brilliantly, others it doesn't — I'll be honest either way.
An estimate of time, costs and potential benefits with stated assumptions. We verify payback after a pilot using actual results.
A prioritised plan — what to implement first, what can wait, and a realistic timeframe for each phase.
An assessment of your IT infrastructure, data and systems. What you have and what you'll need for a successful AI rollout.
Independent recommendations of specific technologies and tools. No vendor lock-in — I pick what's best for you.
Who the AI audit is for
You see potential in AI but don't know where to begin. The audit identifies suitable opportunities, assumptions and risks to guide the next step.
Tried AI and it didn't work? The audit reveals why and maps the path to doing it right. The problem is often the process, not the technology.
Need a solid basis for an AI investment decision? The audit delivers concrete numbers, scenarios and recommendations for your board or leadership.
Need to assess the technical feasibility of AI projects? The audit includes a technical analysis of your infrastructure and integration options.
What you get
📄 Documents
- Audit report (process map + AI opportunities)
- Benefit and cost estimates for suitable opportunities
- Implementation plan with priorities and milestones
- Technical requirements specification
- Executive summary for leadership
🤝 Consulting
- Kick-off workshop with key stakeholders
- Results presentation with Q&A
- 30 days of follow-up support (e-mail, phone)
- Option to continue straight into implementation with me
Sound interesting?
Describe your problem — I'll design a solution and calculate the ROI. The 30-minute call is free and non-binding.
What an audit may reveal
These are illustrative scenarios, not results from particular clients. Savings and payback can only be calculated from your own data.
Online store
Problem: Support repeatedly answers questions about order status and stock availability.
Possible pilot: An assistant connected to the store, with complex cases handed to a person.
Measure: correctly resolved questions, team time and running costs.
Technology company
Problem: Employees ask colleagues about processes (leave, reporting, org chart) and hunt for information in scattered places.
Possible pilot: Search across internal documents with source links and access controls.
Measure: time to a verified answer, accuracy and repeat questions.
Law office
Problem: Lawyers spend time on administration, such as gathering materials and preparing documents.
Possible pilot: Document classification and brief preparation with lawyer review.
Measure: time per case, corrections and review costs.
Company website
Problem: Too few relevant enquiries from the website, but no repeatable process with a clear AI saving.
Possible conclusion: Improve the offer and website measurement first. Add AI when there is a suitable process.
Measure: relevant visits, enquiries and their quality.
Frequently asked questions
Did not find your answer? Try the full FAQ.
An audit of one process may take around 1–2 weeks; a wider audit takes longer depending on scope and access to information. You choose an available slot for the free 30-minute intro call in the booking calendar.
No. The audit is independent — you get complete documentation (report, implementation plan, technical specs) you can act on yourself or with any vendor. Of course, I'm happy to help with implementation too.
Then I'll tell you straight. It's far better to pay for an audit and find out than to put an order of magnitude more into a solution that never pays off. Honesty is the foundation of my work — I have no interest in selling things that don't earn their keep.
Ideally: 2–3 people from management and 2–3 people from the operational team (the ones actually doing the work). Together we map the processes and stay on the same page.
At minimum: your team (headcount, roles), their typical tasks (what gets done weekly), and the systems you use (CRM, ERP, e-commerce). No sensitive company data — just enough context to map the processes.
The initial 30-minute consultation is always free and non-binding. The price of a paid audit with ROI analysis and an implementation plan depends on company size and the number of processes we map — you get a fixed figure right after the intro call.
I map the AI systems you use and flag areas for legal review, especially transparency (from 2 August 2026) and AI literacy (from February 2025). A technical audit is not a legal opinion. For context, see the articles on the AI Act and AI literacy.
Yes. An AI project audit evaluates whether a running (or stuck) AI solution delivers what it should — cost vs. actual benefit, output quality, security and areas to assess under the AI Act. The output is an independent recommendation: fix, rebuild, or stop. Also useful before taking over a project from another vendor.
Mapping your competitors' publicly visible AI use — website chatbots, automated communication, AI content in marketing — and comparing it with your company. You'll learn where you lag, where you lead and what's worth catching up on first. The analysis uses publicly available information only.
Not every company should start with implementation
Sometimes the biggest value is stopping a bad idea in time. In this article I describe situations where AI won't help and where it's better to start elsewhere.
Read about AI's limitsStart with a free 30-minute consultation
No commitment. You describe your situation, I'll tell you whether and where AI makes sense.