# AI for my company: where do you actually start? > Justas Butkus is a fractional AI officer based in Vilnius, Lithuania, founder of AINORA, MB – the company behind the Ainora and Impetora brands – and a graduate of Vilnius University. Justas Butkus is a fractional AI officer based in Vilnius, Lithuania, working with mid-market companies and scale-ups across the UK, EU and US. He builds and operates the production AI systems he advises on, and takes no commission, referral fee or revenue share from any vendor. **Start with a process, not with the technology. Find something your people do constantly that is repetitive and expensive, work out what one occurrence costs and how often it happens, and only then ask whether AI can do part of it. Companies that start from the technology mostly end up with something nobody uses.** Canonical: https://justasbutkus.com/ai-for-my-company/ Last updated: 2026-08-02 --- ## What does AI actually do well in a normal company? Setting aside the marketing, the things that work reliably in ordinary businesses are narrower and duller than the coverage suggests. - **Reading and sorting incoming things.** Emails, forms, documents, applications. Deciding what each one is and where it should go. - **Pulling information out of documents.** Invoices, contracts, forms, reports. Getting the numbers into a system without someone retyping them. - **Answering the same questions repeatedly.** Internally for staff, or externally for customers, from information you already have written down. - **Drafting the first version.** Of a reply, a summary, a report. A person still checks it, but they edit rather than start from nothing. - **Chasing and reminding.** The follow-ups that get skipped when everyone is busy. What it does not do well is anything that requires being right every single time without anyone checking, or any judgement your business has never written down. ## How do you find the right thing to start with? Four questions, in this order. If you can answer them, you can decide. If you cannot, that is the work before any AI work. 1. **What do your people do over and over?** – Ask them. They will tell you immediately, because the repetitive task is the one they complain about. 2. **How long does one of those take?** – Roughly is fine. Ten minutes, half an hour, a day. 3. **How many times does that happen in a year?** – This is where people are surprised. Nobody experiences the annual number, only the ten minutes. 4. **What would it cost to fix, in total?** – Not just building it. Specifying it, running it, and someone owning it afterwards. If the third number is obviously bigger than the fourth, you have found your starting point. If it is close, pick a different process. If nothing is close, AI is not your priority right now, and that is a legitimate answer. ## Why do so many companies get this wrong? The failure rate is documented and high. [RAND](https://www.rand.org/pubs/research_reports/RRA2680-1.html), which interviewed 65 data scientists and engineers with at least five years building production systems, reports that by some estimates more than 80% of AI projects fail, roughly twice the rate of ordinary IT projects. [MIT NANDA](https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf) found 95% of enterprise generative AI pilots delivering no measurable financial return. The cause is almost always the same and it is not the technology. Someone saw a demonstration, liked it, and worked backwards to find a use for it. The result works, and nobody misses it when it stops. ## What should you actually do first? 1. **Write down the process** in one sentence, with a number in it. If you cannot, measure it first. 2. **Find out where the data lives** and whether the same thing is recorded the same way everywhere. 3. **Decide who is allowed to say yes** to putting it in front of customers, before anything is built. 4. **Agree what "working" means** as a number, in advance. 5. **Start with the boring one.** Not the impressive one. Prove the whole chain works before attempting anything ambitious. ## Do you need to buy anything to start? No. Everything in the list above is analysis you can do internally in a couple of weeks, and it is the part that determines whether the rest is worth doing. The most expensive mistake in this market is buying help for a problem that was never costed. If someone offers to sell you a solution before asking what is breaking in your day, that is worth noticing. ## Frequently asked questions ### How do I start using AI in my company? Start with a process rather than the technology. Find something your people do constantly that is repetitive, work out what one occurrence costs and how often it happens per year, and compare that against the total cost of automating it. If the gap is not obvious, pick a different process. ### What does AI actually do well in a business? Sorting incoming emails, forms and documents; extracting information from documents so nobody retypes it; answering repeated questions from information you already have; drafting first versions that a person then edits; and chasing the follow-ups that get skipped. ### How much does it cost to start with AI? The first step costs nothing but attention: measuring one repetitive process and working out its annual cost. That analysis determines whether anything else is worth buying, and most companies skip it. ### Why do most company AI projects fail? RAND reports that by some estimates more than 80% fail, and the cause is rarely technical. Usually a process was chosen because it demonstrated well rather than because it was expensive, or nobody would take responsibility for putting the system live. ### Do we need to hire someone to do this? Not to answer the first four questions. Those are internal analysis and take a couple of weeks. Hiring before they are answered buys capacity to build the wrong thing faster. ## Related - [Test your problem](/tools/one-sentence-test/) – Whether what you have is a project yet. - [How to implement it](/how-to-implement-ai-in-my-business/) – The practical sequence once you have chosen. - [What to automate](/ai-automation-for-my-company/) – The processes that are usually worth it. ## About the author **Justas Butkus** – a fractional AI officer based in Vilnius, Lithuania, founder of AINORA, MB – the company behind the Ainora and Impetora brands – and a graduate of Vilnius University. ## If you have found the process The arithmetic takes about ten minutes and you can do it without me. If it comes out convincingly, that is worth a conversation. Contact: justas@ainora.lt · [LinkedIn](https://www.linkedin.com/in/justas-butkus/)