One integration at a time: how to start with AI without a big-bang project

There are two ways a warehouse or distribution company can bring AI into its operation. One is the big-bang project: a programme with a steering group, a long list of use cases and a go-live date a year away. The other is one integration, built properly, used by real people, and then the next one.

We have watched both. The programme usually produces documents. The single integration usually produces a change on the floor within a few months, and a clear idea of what to do next.

Why one well-built integration beats a wide programme

A programme spreads a fixed amount of attention across many things. Each use case gets a little design, a little data work and a little testing, and none of them gets enough to survive a busy week. The floor sees nothing for a long time and quietly concludes that AI is a head office hobby.

One integration concentrates that attention. The data it needs is checked properly. The exceptions it meets are written into the checks. The people who use it are the ones who were in the room when it was designed. When it holds up through a peak, it earns the trust that the second integration will need, and it teaches everyone what the next good candidate looks like.

It is also easier to stop. If the first integration turns out to be the wrong one, you have lost a few weeks, not a year.

The seven steps

Here is how we work through one integration, in order. Some steps we do together; some redfly does and brings back to you.

1. We work out the requirements together. You show us how things work today: the shift, the systems, the paperwork, the places where people re-key, re-check or wait. We agree on the first integration from that.
2. We design it. redfly decides where the AI runs, what it reads, what it hands back and what happens when the connection drops, and writes it down.
3. We review it together. We present the design to you, your supervisors and your IT contact, and refine it until it matches how the floor really works.
4. We build it. redfly builds the integration against your systems, through the warehouse system's API or a copy of its database, the mailbox, the sensors or the cameras, whichever it needs.
5. We pilot it together. A small group uses it on real work: one shift, one client or one dock.
6. We test it together. We test it with a set of checks, agreed in advance, until it holds up on a busy floor, including the strange inputs the pilot turned up.
7. We roll it out together. It goes live for everyone, with our team watching what comes through until your people no longer need us to.

Nothing here is unusual. It is how good software has always been put into a working building. What matters is that each step is done fully before the next begins, and that the floor is involved from the first step rather than shown the result at the last.

What you bring and what redfly brings

The customer side of this is simpler than people expect. You do not need a data strategy, a cleaned-up database or a view on which AI model to use. You bring how you work: the people who do it, the systems they use, the paperwork they handle and the honest account of where it goes wrong. One contact on your side is enough to start.

redfly brings the rest: judgement about where AI fits and where it does not, the engineering to connect it to the systems you already run, the discipline of checks and pilots, and experience of keeping software working at busy sites with poor internet. Our founder spent years building the software behind large warehouses in the US and Europe, and that is the background the design draws on.

Where the results land

An integration is only useful if its output reaches someone who can act on it, on a screen they already look at. That is what the redfly Portal is for. Alerts, such as a temperature drift or a late order risk, suggestions, such as a dock slot or a stock move, and dashboards all land there, for your supervisors, dispatchers and customer service people to review and act on.

Where the integration writes into your own systems, such as an order entered from an email, it does so through those systems' normal doorway, and the redfly Portal shows what was done so a person can see and check it. A person stays in charge throughout.

Then the next one

Once the first integration has been through a peak, the second is usually obvious, because the people using the first will tell you. The requirements step is shorter, the design reuses what was learned about your systems and your link, and the checks from the first pilot become the starting point for the next.

Start with the one integration a supervisor would miss if it stopped; the second one picks itself.

This is the shape of how redfly works with warehouse and distribution companies: one integration at a time, in these seven steps, woven into the systems you already run, with the results landing in the redfly Portal for your people to act on.