We move
→
meets the unknown
…into a
direction
A guarded edge, not a wall. more ↓
We make the data you have — and the data you didn't know you could have — speak, and turn it into decisions and time. Inside: your company's data, and the processes someone still redoes by hand every week. Outside: public sources nobody reads systematically.
In plain words: we analyse business data to guide decisions, we automate repetitive processes to give people back the time for what matters most, and we aggregate public data for a range of industries. We work with companies, professionals and public bodies.
A British company, with an Italian presence. Incorporated in London, where Europe's AI is built; operating in Italy, where it has to work.
A guarded edge, not a wall. more ↓
Limes was, in ancient Rome, the frontier line of the Empire: the point where the known met the unknown. From it come limit, liminal, preliminary.
We work where AI meets what it cannot do yet. The limes doesn't separate to exclude: it marks the point where pushing further is worth it.
A useful AI is not one without boundaries, but one that knows where they stand: guardrails, context, accountability. The limit is design, not obstacle.
"Limes" sounds like lime, the citrus: sharp, fresh, clean-cut. Latin rigour and contemporary lightness in a single gesture.
"We don't hand you a scraping tool. We move your data-acquisition boundary — with a measurable result and a date."
Prototypes in days, not in months. Shorter time to deployment without giving up technical depth.
The limit as a design measure. Every solution starts from where the boundary sits — and from how far it pays to move it.
The AI we build is not a black box running loose: it is a supervised tool, with guardrails and accountability.
Analysis of your business data to guide decisions: sales, customers, inventory, costs. Systems that often don't talk to each other, turned into numbers you can decide on — not a dashboard nobody opens.
Qualitative analysis and aggregation of public administration and institutional data, for a range of industries. Information published by law that almost nobody reads systematically: we make it queryable and turn it into opportunity.
Business process automation: the repetitive tasks that keep good people busy. The goal is not to replace them — it's to free them for the higher-value work only they can do.
Three different fronts, always approached the same way.
We look where nobody's looking. Public administration portals, open databases, institutional sources that publish by law and that almost nobody reads systematically: we collect them, normalise them and make them queryable.
The boundary drawn with precision. Automations built on the process you actually have, not generic modules forced to fit. A system you trust is a system you know where it stops.
Working prototypes in days. You use them with your real data and see at once whether the course is right, before investing quarters. Speed as method, not shortcut: we go to production fast because we know where the risks are.
We have worked with networks and data since 1989, when the Internet came through academia and research, years before a provider market existed. The craft has stayed the same across seven technology transitions: network access, infrastructure, CRM and customer data, e-commerce, the governance of large data estates — MDM and PIM — up to today. Connecting systems that don't talk, extracting usable data, turning it into a decision you can measure.
In the Nineties: corporate networks, websites and consumer Internet access, as a provider, when getting online was still a craft and not a given. From 2001, consumer data and loyalty programmes for major brands — when customer data was becoming an asset to govern, not an archive to fill.
Online channels for major retail chains, built when they were still a gamble and brought to scale: from the first million in sales in 2011 to over four million a year in 2017 — 16% growth achieved while cutting half a million in costs, not adding it. And Christmas peaks run as an operation, not suffered as an emergency.
Master data management and PIM in a large multi-banner retail group: 5.65 million customer records deduplicated from eleven sources, and a business case with payback between 12 and 25 months, in the GDPR-compliance window.
Artificial intelligence is the seventh transition we cross. And we cross it like the other six: without falling in love with the tool, keeping our eyes on the process it has to improve.
A crew of fifteen specialists who have worked together for twenty-five years, across projects from different technology generations. Not a group assembled for a single engagement: people who know each other, and who have already crossed more than one paradigm shift together.
The experience told on this page belongs to the people who form Limes AI, built over their professional careers. It is the baggage we work from: method, measurement and data governance are not improvised on one project — they accumulate over thirty years of projects.
The customer database of a large retail group, fed by eleven sources that didn't talk to each other — websites, paper forms from several banners, apps, social, Excel files, suppliers — brought back to a single record with quality rules and change tracking.
Boundary moved →from eleven partial truths to one customer, recognisable across every banner.Learn more ↗
A master data management project presented with an estimated annual saving and a payback between twelve and twenty-five months, in front of an external assessor sent by the shareholder. Two years later, the same data asset was generating direct revenue.
Boundary moved →from cost centre to revenue centre, with the numbers written down first.Learn more ↗
A retail chain's online store taken past one million euros in annual sales, run on a full P&L down to EBITDA, signed off by the person in charge — not on a traffic dashboard.
Gate opened →from the channel that "does fine" to the channel with a margin, measured every month.Learn more ↗
We start from the exact point where you stop today: the manual process, the data you can't find. Where it's worth pushing further — and where it isn't.
Learn more ↗In a few days you have a working artefact, not a slide. You get your hands on it, and we correct the course together before investing months.
Learn more ↗Context, limits, quality controls, accountability: the tool becomes reliable and repeatable. Not a black box.
Learn more ↗We deliver a result, not man-hours: a boundary moved, measurable with a number and a date. And it scales the way it started — first a trial on a small perimeter, then the extension, never the other way round.
Learn more ↗A project done well does not create the privacy problem: it formalises and governs it. And it is not a position we took when AI arrived.
2005–2009: in a children's brand digital club, it was the parents who registered their kids, with adult and child data kept separate. Consent didn't need proving: it was built into the account architecture. Twelve years before the GDPR named that principle privacy by design.
2013: database ownership with the client, right of extraction at any time, usage limits for the supplier. It is the clause negotiated with AI vendors today — negotiated five years before the GDPR.
2017: an external supplier is given access only to the subset of data within its remit, with dedicated quality rules. It is the question customers ask about AI today, solved eight years earlier on a supplier portal.
Digital is not a project — it is a change in the way of thinking. Written and signed in 2017. Today we say it like this: AI is not the project — the project is your business process. That is why privacy is not a clause added at the end: it is an architecture decision taken at the start. Let's talk →
Tell us in two lines. We will tell you by how much — and how fast — we can move it.
The data you leave is used to reply to you. How we handle it is set out in the privacy notice.
office@limesai.uk · +44 20 3807 7897
Limes AI LTD — London