AI implementation, step by step.
What it takes to get AI from a good idea to something your team uses every day: each step, how long it usually takes and what we need from you.
Book a free 30-minute callWhat is AI implementation?
AI implementation is the work of taking AI from an idea to something your team uses every day. It means choosing the right job for it, connecting it to the software you already have, testing it on real work, showing your team how to use it and keeping it working after launch. Most of the effort is in the business, not the AI.
Seven steps, from idea to in use.
1. Talk it through
A free 30-minute call on Google Meet about how your business runs, where the time goes and what you’d like to be easier. If there’s a sensible next step, we’ll suggest it.
What we need from youHalf an hour, and a rough idea of what’s taking too long. Nothing to prepare.Typically30 minutes2. Find the right job
A Fit check. We talk to the people doing the work, look at the tools, spreadsheets and messages you use, and find where AI and automation would genuinely help. We also say where they wouldn’t.
What we need from youTime with the people who do the work, and a look at the tools involved.TypicallyAgreed with the price, depending on how many people and tools are involved3. Agree a written plan
A plain-English plan: what we’d build, which tools it connects to, who checks what, how we’ll know it’s working, and a price. There’s no commitment to build.
What we need from youSomeone who can make the decision. You can act on the plan with us, on your own, or not at all.TypicallyAs long as you need to decide4. Build around your tools
We set up and connect the software you already have, including AI you may already pay for like Microsoft Copilot, before adding anything new, and build what the plan says is worth building. A person approves anything that goes to a customer.
What we need from youAccess to the tools involved, or someone who can grant it, and one person to answer questions.TypicallyUsually two to six weeks for one job across one or two tools; longer, in stages, for more5. Test it on real work
We try it against real examples from your business, not a demo. Where it helps, it runs alongside the old way until everyone trusts it. If it isn’t good enough, it doesn’t go live.
What we need from youReal examples, a few people to try it, and honest feedback.TypicallyUsually one to two weeks of normal work6. Go live and train the team
We switch over, show your team how it works and what to do when something looks wrong, and leave short written notes so new starters can pick it up.
What we need from youTime for your team to learn it, and a say in when to switch over.TypicallyA day or two of training, spread to suit your week7. Keep it running
The Run stage, if you want it: we monitor it, fix it when a connected tool changes, and improve it as your business changes around it.
What we need from youTell us when something in the business changes, such as a new tool, price list or process.TypicallyOngoing, for as long as you want it
Why AI projects stall, and how we avoid it.
The common reasons AI never makes it into everyday use. Most have nothing to do with the AI itself.
Starting with the tool, not the job
Someone buys an AI subscription and looks for a use for it. It gets tried, then forgotten.
How we avoid itWe start with how the work actually runs, and only suggest AI where it earns its place.
A first project that’s too big
The plan tries to change everything at once, takes months, and nobody can tell whether it’s working.
How we avoid itWe start with one job that’s used often and easy to check, then build from there.
Information that’s scattered or out of date
The AI is only as good as what it reads. If job details live in five places, it gets things wrong.
How we avoid itWe check this in the Fit check, and either tidy the basics first or pick a job that doesn’t depend on them.
The team isn’t involved
It’s designed by people who don’t do the work, so it doesn’t fit the real job and people work around it.
How we avoid itWe talk to the people who do the work, test with them and change it based on what they say.
It never leaves the demo
It looks good on sample data, but it lives in a separate app nobody opens, or breaks on real examples.
How we avoid itWe build inside the tools you already use and test on real work before anyone relies on it.
Nobody looks after it
A connected tool updates, something stops working, and no one notices until a customer does.
How we avoid itThe same people who built it monitor and fix it in the Run stage, so it doesn’t quietly break.
What we need from you.
Less than most people expect. We’ll be specific about it in the written plan.
Time with the people who do the work
A few conversations with the people who do the job day to day, not only the people who manage it.
Access to the tools involved
Logins or permissions for the software the job touches, which you grant and can take back.
One person who can decide
Someone who can answer questions, agree changes and say when it’s ready to go live.
Real examples
Real jobs, emails, documents or orders to test with, so we know it works on your work, not a demo.
How long it takes.
Typical ranges, not promises. Your written plan gives timings for your job, and the cost is explained on How we work.
- 1The introductory call: 30 minutes, free.
- 2A Fit check: agreed with its price, so you know when to expect the plan.
- 3A first build, one job across one or two tools: usually two to six weeks.
- 4Bigger work across several tools or teams: a few months, delivered in stages so the first part is useful early.
- 5Testing on real work: usually one to two weeks before your team relies on it.
What makes it take longer
- More tools to connect, or tools that are hard to connect to
- Information spread across inboxes, spreadsheets and paper
- More people who need to agree, learn or change how they work
- Waiting for access from a software supplier
What we implement.
You might need one or a mix. The Fit check works out which, and in what order.
- AI assistants and agentsAI that drafts, sorts and summarises routine work, and agents that carry a task through several steps.
- Workflow automationThe routine steps between your tools happen on their own, so details are entered once.
- Custom AI softwareDashboards, planners and internal tools built around how you work, with AI where it helps.
- Ongoing supportWe monitor, fix and improve what we’ve built, and train your team, so it keeps working.
AI implementation questions.
More in the full list of questions.
What is AI implementation?
AI implementation is the work of taking AI from an idea to something your team uses every day. It means choosing the right job for it, connecting it to the software you already have, testing it on real work, showing your team how to use it and keeping it working after launch. Most of the effort is in the business, not the AI.
What does an AI implementation partner do?
An AI implementation partner finds where AI would help your business, builds it around the software you already use, and keeps it working afterwards. With us, that’s one offer in three stages: Find (a Fit check), Build and Run. The same people do all three, so nothing gets lost in a hand-off.
How long does AI implementation take?
It depends on how many tools and people a change touches. As a guide: the Fit check’s timing is agreed with its price, a first build for one job across one or two tools usually takes two to six weeks, and bigger work takes a few months, delivered in stages. These are typical ranges, not promises; the written plan gives timings for your job.
What does AI implementation cost?
It depends on the size of the job, and every stage is priced in writing before work starts. The 30-minute call is free. After it, we tell you what a Fit check would cover and its price, and the Fit check prices anything worth building. Ongoing support is optional and priced separately. Any software subscriptions or AI usage charges are paid to the provider, and we point them out up front.
What’s the difference between AI consultancy and AI implementation?
AI consultancy usually ends with advice: a report or a plan. AI implementation ends with something your team uses. We do both: the Fit check is the advice, and Build and Run are the implementation. You can stop after the plan if you’d rather do the rest yourself.
Do we need an AI strategy or clean data before we start?
No, and many small businesses don’t have either. The Fit check looks at how your work runs and where your information lives, and the plan says what needs tidying first, if anything.
Will we have to replace our software?
In most cases, no. We build around the tools you already use, and only suggest something new where it would clearly help.
What happens if it doesn’t work as expected?
We test it on real examples from your business before your team relies on it, and often run it alongside the old way first. If it isn’t good enough, it doesn’t go live: we change it, or tell you plainly that it isn’t worth pursuing. You decide after each stage whether to go on.
Is AI implementation worth it for a small business?
It can be, when it takes a job your team does often and makes it quicker or more reliable. It isn’t worth it for a job that happens once a month, or where better setup of the software you already pay for would do. We’ll tell you which is which.
Who checks what the AI does?
A person approves anything that goes to a customer. AI handles the routine parts, and your team makes the decisions that need judgement.
Who we help.
Businesses where an office keeps work moving on site, on the road or on the shop floor. If yours is different, the call is still the place to start.
Read next.
- Propel, the AI implementation partnerWho we are and the one offer: Find, Build and Run.
- How we workThe three stages in detail, and how the cost works.
- The Fit checkThe first stage: a written plan and a price for anything worth building.
- AI strategyDeciding where AI fits in your business before anything is built.
- AI integrationConnecting AI to the software you already use.
- AI consultancyAdvice from people who also build it and keep it running.
Book a free 30-minute call.
On Google Meet, with Jacob. Tell us about one job that takes more effort than it should, or how a normal week goes.
Pick a timeOr email jacob@propelai.co.uk