Integration & Automation
How your systems talk to each other, and what stops being done by hand.
Integration architecture and the automation built on it — designed before a connector is opened.
Implementation. Delivered for you, or built alongside your team so they can carry it on.
Volume is rising faster than the team, the same data is being keyed into two systems, or an improvement program has stalled at the mapping stage.
The practice
Process discovery and design
The work captured as it actually runs, including the steps that live in someone's head, and redesigned before anything is automated.
Integration architecture
How the systems talk: which one owns each record, what the contract between them is, what is real time against what is a nightly batch, and what happens to a message that fails at two in the morning.
Systems integration build
APIs, connectors, queues and file interfaces built against that design, with retries, replay and alerting, so the integration is monitored rather than assumed.
Platform selection
Which of Power Platform, Azure Integration Services, UiPath or an agent the work actually needs — decided against the process, not a preference.
Intelligent processing
Where inputs vary too much for rules — documents in forty layouts, email that has to be understood before it is routed — an agent takes that part.
Master data and identity
Which system is the source of truth for a customer, a site or a part number, and how the others are kept in step. Most integration failures are really this question, unanswered.
Measurement
A baseline before the build and the same numbers after it: touch rate, cycle time, rework, backlog.
Adoption and change
The people around the process trained, the exceptions owned, and the old way closed off. Automation that nobody adopts is a licence with a story attached.
Building your capability
Where you want this to become an internal muscle, we build with your team and teach the pattern rather than leaving a black box behind.
Operate or hand over
Documentation, rules written down and a named owner on your side. If you would rather we kept running it, that is a conversation.
Worth an afternoon before they are worth a platform.
Most integration estates were not designed, they were added to — point to point, by whoever needed it, with credentials in a flow somebody has since left behind.
Settled before a connector is opened, because every one of these is expensive to change once traffic is flowing through it.
- Who owns the record
Which system is the source of truth for a customer, a site, a part number — and what the others may do with it.
- What the contract is
The fields, the format, and what happens when one end changes them without telling the other.
- Real time or batch
Most things asked for in real time are fine hourly, and cost a fraction as much to run.
- What happens when it fails
Retry, dead-letter, replay and alert. The integration nobody thinks about is the one quietly dropping messages.
- How you know it is working
Monitoring that tells an owner before it tells the business.
We map the process as it runs, design the integration architecture around it, and build: Power Automate and Power Apps, UiPath where a system has no API, Azure Integration Services where the volume is real, and an agent where the inputs vary. Measured before and after, and handed over with a named owner.
The question is not where AI could go
It is which work should not be done by hand at all, and what that work actually needs:
- Unambiguous steps are rules, and plain automation runs them more cheaply and breaks less often than anything with a model in it.
- Two systems that will not talk need an integration, not an assistant.
- Inputs that vary need an agent to interpret before anything can be actioned.
- A spreadsheet that has become the system of record needs an application.
Working out which of those you are looking at is the first part of the job.
Integration architecture, before anybody opens a connector
Most integration estates were not designed. They were added to. Every new connection was point to point, built by whoever needed it, with credentials in a flow somebody has since left behind. It works until a system is upgraded and six things break, none of which are documented.
The design questions are the same every time, and they are worth an afternoon before they are worth a platform:
- Who owns the record. Which system is the source of truth for a customer, a site, a part number — and what the others are allowed to do with it.
- What the contract is. The fields, the format, and what happens when one end changes them without telling the other.
- Real time or batch. Most things that people ask for in real time are fine hourly, and cost a fraction as much to run.
- What happens when it fails. Retry, dead-letter, replay and alert — because the integration you never think about is the one quietly dropping messages.
- How you know it is working. Monitoring that tells an owner before it tells the business.
Get those right and the tooling is almost incidental. Get them wrong and no amount of platform licensing saves you.
Which platform, and why
The choice is made against the work, and each one has its own page: Power Platform where the process lives in Microsoft 365 and a person still has a step in it, Azure Integration Services where integration volume is real and systems have to be decoupled, UiPath where a system has no API and the only way in is the screen a person uses, and an agent where the input varies too much for rules.
We will say when the cheapest answer is the least interesting one, and when the honest answer is that the process should be fixed before anything is automated.
One at a time, because results buy the next one
A program that improves everything slightly is impossible to defend at the next budget conversation. A queue that took four days and now takes one is not.
What changes
A finance team taking supplier invoices by email in a dozen formats, keying them into the ERP, matching them against purchase orders and chasing the differences. Afterwards the copying is gone: documents are read on arrival, matching runs inside an agreed tolerance, and anything outside it reaches a person with the reason attached.
Work we have done
Data integration hub — commodity trading and risk. Inventory, pricing and related sources connected into one place so the business could analyze current data rather than stale extracts, implemented on Posit. Most of the effort went into environment access and the connection to each source rather than into the analytics. More on that engagement.
This, or data integration
This page is systems talking to each other: an order moving between two applications, a document arriving and being actioned, a process that currently runs on copy and paste. Where the need is data landing somewhere for reporting and analysis — pipelines, transformation, freshness — that is data integration, and it is scoped from the decisions the numbers serve.
Have a process or an interface in mind? Bring it to a briefing.
Not sure which work is worth automating? The assessment answers that in three weeks.
What you keep
The code and the data·Your existing relationships·Approval and control·The ability to stop·The off switchWhat that means
The tool follows the problem.
We build on what you already own wherever that works, because a licence you are paying for is cheaper than a platform you are about to buy.
- Power Automate
- Rules, orchestration and approvals inside Microsoft 365, where most of this work already belongs.
- Power Apps
- The interface for the step a person still has to take — an exception queue, a review screen, a form that writes back.
- UiPath
- Robotic process automation for systems with no API, where the only way in is the screen a person uses.
- OpenAI and Anthropic
- Interpretation: documents in dozens of layouts, email that has to be understood before it can be routed.
- Copilot Studio
- Agents published where people already work, when the process lives inside Microsoft 365.
- Azure Integration Services
- Logic Apps, Service Bus, API Management and Event Grid where integration volume is real. The practice.
- Your line-of-business systems
- The ERP, finance system or operations platform the work already touches, connected rather than replaced.
Technology Strategy & Advisory
Strategy tied to what the business is trying to do, and the advisory weight to act on it.
Learn more ↗Data & Analytics
What to do with the data estate you have, and the platforms you are already paying for.
Learn more ↗Training
Process, automation and AI — taught to the people who have to do the work.
Learn more ↗AI Agents
Deciding where an agent earns its place, and building the one that does.
Learn more ↗Start with a 45-minute briefing.
No pitch. Bring the decision you are stuck on and we will tell you what we would do about it, including when that is nothing.