Skip to content

Integration, data and practical AI

Connect the systems you have. Then make the data useful.

Your numbers are in spreadsheets, your documents are in inboxes and your systems do not talk to each other. The fix is usually integration and a reliable data foundation first, then reporting and automation. AI comes in where a specific, checkable job justifies it, with a person still responsible for the decision.

What Blacklight delivers

Data platforms and integration

One version of the truth, not truth scattered across tools that do not talk. We connect the systems you already run: SAP, Oracle, finance and housing platforms. We centralise the data where you can trust and audit it. We did that for Sulzer: 88 engineering sheet types standardised into one central system used across 150 countries.

Dashboards and reporting

Reporting that people open: Power BI dashboards and case queues built on live data, like the real-time dashboards we delivered for Wandle Housing Association's customer service teams. If your Power BI need sits inside a wider Dynamics or Dataverse estate, start at our Power Platform page.

Document processing and automation

The plain work with the fastest payback: the system reads, routes and validates documents, so people stop re-keying them. For Stonewater, Azure AI reads incoming invoices and validates them against the financial system. It flags exceptions or processes the invoices automatically under business rules.

Applied AI: where it pays for itself

AI applied to a specific, measurable job: classify documents, match records, flag exceptions. For dotDigital we built a fuzzy-matching engine that scores incoming leads against existing records: new, duplicate, or auto-merge. Machine judgement now does a job humans did badly in spreadsheets.

The honest part about AI

Most of what you read about AI is written by someone who sells it. This is the version we give you in a scoping call:

Where it pays off today: reading and validating documents at volume, matching and de-duplicating records, flagging exceptions for a person to decide, and answering questions over your own well-organised data. The common thread: a narrow job, high volume, and a person who owns the important judgement calls.

Where it does not yet: any task where a confidently wrong answer is expensive and no one checks. If your data is scattered and inconsistent, AI amplifies that problem before it repairs anything. This is why most failed "AI projects" were data-platform projects that got skipped.

The order that works: first sort the data foundation, second automate the obvious workflows, third apply AI where a measurable job justifies it. If you ask us for AI and you need a data platform, that is what we quote for.

Proof

Automation that survived contact with the finance system

  1. Problem

    Stonewater is a housing association that manages about 33,600 homes. Its documents were dispersed across repositories. Search was unreliable. Invoice handling ran on manual effort.

  2. Build

    A SharePoint document-management platform that applies metadata automatically. A standalone .NET supplier portal. Purchase-to-pay processing where Azure AI reads each incoming invoice, validates amounts and work-order codes against the financial system, and flags exceptions or triggers payment under business rules.

  3. Result

    Tens of thousands of documents are in one structured environment. Suppliers manage their own submissions. The system validates invoices before approval, without manual filing. Subject Access Request compliance is complete across the document estate.

Where data work meets the rest of the estate

Data work rarely arrives alone. If the answer includes a system that does not exist yet, such as a portal, a line-of-business build or an integration layer, that is our bespoke practice: the same team, in the same building. If your estate is already Microsoft, with Dynamics 365, Dataverse or Power BI, the Power Platform page shows how we work inside it.

Bespoke software development →

Power Platform →

Frequently asked questions

We are still on spreadsheets. Where do we start?

Start exactly there. The scoping call starts with where your data is today, not where a vendor wants it to end. The usual first step is small: one workflow, one report, or one document process that proves the value before you commit to a platform.

Where does our data live?

Where your data lives is a design decision, agreed with you and written down before the build starts. For Microsoft solutions that is usually your own tenant, in the region you choose. In scoping, we tell you where every component runs and which services can see which data.

How do you keep AI under control?

We scope the job narrowly and keep a person in the loop where it matters. In the Stonewater build, AI validates invoices against the financial system and flags exceptions to people. It does not guess unsupervised. That pattern is our default: automate the volume, escalate the judgement.

Do we need AI at all?

Possibly not yet. If the data foundation is not there, a platform-and-reporting project usually returns more, faster. Scoping tells you which one you are in.

Talk to us

Tell us where your data is.

Thirty minutes, with no pitch deck. You will leave with a view of what is trapped, what it takes to free it, and whether AI belongs near it yet.

Get in contact