Integration Love Story med Tom Canter

Introduktion

Tom, thank you so much for joining Integration Love Story. We are really glad to have you. As always, our first question is: can you please introduce yourself to our viewers and listeners?

I’m Tom Canter. I’ve been on and off on Microsoft VTS — the Virtual Tech Specialist program — which got me kicked off in the Microsoft-centric integration world. But my primary focus has been integration since I came aboard at Neon in 1998, and I’ve been doing integration throughout that time. I’ve seen a lot of interesting things over the years.

Something I want to say, as a small plug for the community: it’s a very customer-centric but also very connected community of people that I trust every day and am honored to be a member of. They work hard in the bowels of the engine and they make things work. What’s great about our community is that it’s well connected and people trust each other. You don’t feel like you can’t share your problems or your issues, and you always get help. There’s always a hand you can take if you need it.

Scubadykning, piano och livet i Seattle

What does the private Tom look like — what do you do outside of integration?

One thing is that I’ve made connections and friends all over the world, and that’s led me to do things like in 2014 becoming scuba certified, because Saravana decided he wanted to hold an Integrate event in Sydney, Australia. I wasn’t going to go to Australia for the first time and not do something like the Great Barrier Reef. So I became scuba certified — advanced open water — and had the trip of my life.

I’ve also played piano since I was seven years old. One of my great comforts and relaxation is to play during the day. I lost my father about a year and a half ago, and through that came my mother’s music collection that she and my father had built together. I’ve been going through and cataloging 4,000 different songs and music books, just revisiting that history.

My wife and I love to travel, and we use integration events as an excuse to go places like Hamburg and London. And of course I have my daughter and grandchildren — they live up in Canada. Watching my grandson learn to play soccer and seeing my other grandson’s drawings, he’s very talented. There’s quite a community around us here in downtown Seattle, where we’re very connected with our neighbors and friends. We’re very fortunate to live in a beautiful little area.

Piano or scuba diving — which is closest to the heart?

They both have their own aspects. When I play piano, I go into a somewhat meditative state where my mind roams freely. When I’m diving, I’m very attentive to the environment but also relaxed, because it’s beautiful and enjoyable. They both have their own part.

Vad som inte förändrats sedan 1998

You’ve been in the integration field and community since the very beginning. What has not changed since you got into integration?

The fundamental premise of integration is that systems that weren’t designed to communicate but should be, are brought together by the skill and effort of people like us who make things work. Why do I have to type this in when I have it over here — why is it not coming to my system automatically? Making parts of systems work together that should, maybe were thought about, but not built that way.

Take healthcare visibility — making that seamless between different applications has a real health outcome. It causes the doctor to be completely aware of all your various conditions and treatments. Integration makes our life more seamless and more complete, because it puts together things that maybe the developer who’s very appropriately focused on solving their own business problem doesn’t see as part of their long-term job.

I was in the US Navy and I was what’s called a snipe — jokingly, a human who never sees the light of day because they work in the bowels of the engine and never come up topside. We’re kind of snipes. We’re digging around and asking: what if I turn this wrench and how will that make things better for everybody? That’s integration. Connecting systems that aren’t necessarily upfront designed to connect.

Lärandet som superkraft

If you should learn one skill as an integrator — besides technical skills — what would that be?

This is going to be an oddball one: the willingness to learn. The open mind to learn about new things, how they work, and dig into them and become a micro-expert. It’s kind of an oxymoron, but I think that’s the case.

It’s a bit like what actors do. They come into a role and they learn very deeply about their part. They may never use it again the rest of their life. As integrators we’re somewhat like that. We come in, we do a role — oh, you’re the data guy today. You’re the web service guy. You’re the security guy today. You wear too many hats. But that’s what makes it fun. Your job tasks change almost every day according to what you’re working on.

Det minnesvärda misstaget

Do you have a memorable mistake from all those years?

I do. We were at a healthcare client, might have been in Louisiana, and it was the day before production go-live. We were running BizTalk at that time. One of my test engineers turned to me and said: “I’m having trouble — I can’t get this database to delete.” I said: “Check that box right there and it’ll force it. It’ll disconnect everything and force it to delete.” He does that. A few minutes later the project manager comes wandering in and says: “I think production is down.”

It just so happened — and it was not his fault, I was standing over his shoulder advising him — that the database he deleted was production, not the test environment.

Fortunately, I had impressed upon them and insisted that they enable all the BizTalk backups. BizTalk is quite unpleasant when it decides to go down. But the database team had restored the backups 15 minutes later and we went live to production the next day.

How long before you knew what had happened?

The minute the project manager walked in the door — not even when he asked the question — I knew.

The first question on any troubleshooting checklist: what was the last thing that changed?

Exactly. And here’s the interesting thing — it’s not always the integration itself that’s the problem. But the integration is usually where you can tell something’s wrong, because the logs and the observability are all in one place. That unfortunately leads to the belief that the integration is the problem, when it’s often the actors around it. But we’re always the ones who find the issue.

Vad vi saknar från BizTalk-eran och observability idag

What I miss from the BizTalk days is that when something went wrong, you went to the logs — the Windows logs, the SQL logs — and you found everything in one place. It’s not impossible with the newer approach, but you have to work harder for it.

The measure of whether there’s a gap is when products appear that try to fill that space — and there are a lot of products attempting to do exactly that on the cloud platforms right now. I hope they’re going to be successful. I want one place where I can gather all the data from the cloud, navigate it easily, and find where things happened.

Then again, if you have all the data in one place that everybody can look at — now we put on the security hat. So there’s that.

Agentic AI och Swiftmate

You were involved in a project called Swiftmate AI. What is that and what problem did it solve?

It’s a funny story about random things coming together. I had time on my hands, and I live in a condominium of 10 units here in downtown Seattle. One of my neighbors mentioned that his brother runs a mitigation business — they go out and fix houses after a pipe burst.

Their biggest challenge was how long it took to get an estimate to the insurance company. The expert goes out with a clipboard, walks around, writes notes. He plans on going back to the office to write it up and submit it. But 30 minutes later, there’s another house with a burst pipe — an emergency — and he goes there instead. Sometimes it might take a week to get around to completing that first estimate. In the meantime, the homeowner can’t get their house fixed until the insurance company comes in and approves the repair.

The question was: can we shorten that time using AI?

They gave us a list of 30 to 40,000 line items to pick from, and a collection of already completed estimates based on transcriptions from the field. We used AI to assess whether the problem was being described correctly and whether the output matched.

We wrote an agentic process that fixed the prompt recursively. An agentic AI ran through a quality assurance manager checking the prompt, a prompt writer expert, a resource manager, a mitigation specialist — all these agents worked together to iterate over the prompt and slowly improve it until we reached about 90% accuracy.

The only resources we invested were Azure AI time and a server sitting on my desk running the agents. We didn’t build a UI or any other infrastructure, because there was no point if the core experiment failed.

It succeeded. It’s out there now.

And what was the business outcome?

In the mitigation industry, the customer is focused on immediate recovery — it’s about who gets the estimate to them first. Now their inspector goes out, comes back to the car, and produces an estimate. Their competitors go back to the office and come back a day or a week later. So it’s not saving them time. It’s driving more business.

Long term, the goal is to build a feedback loop. The expert will come back, make corrections to the estimate before submitting it to the insurance company, the insurance company will give feedback, and that all feeds back into improving the prompt over time. It becomes a continuously learning agentic process.

Logic Apps och agentic AI — verktyg i ett nytt sammanhang

What we’ve done as integrators — creating tools that move data from one system to another — those are exactly the tools you give to an agent. You’re giving it a toolbox. And when you give good tools to a smart agent and give it instructions, suddenly your integrations have even more value than before. You don’t need to change a lot. And in many of these situations, it gives you the ability to do things that were practically impossible before. It lets you attack problems that were intractable.

How should you approach agentic AI if you want to find the right problem to solve with it?

Go back to the minimum viable experiment approach. Ask yourself: what do I want it to do? What might solve that? Then design an experiment that proves out the core concept, keeping it as lightweight as possible. Once you’ve proven the core concept, you know whether it’s worth going further. If you keep the scope small and the cost low, you’re not afraid of attempting it. And then you build from there.

Kyra — ett hackathon-projekt om Azure-arkitektur

I also worked on a project called Kyra, which came out of a hackathon. The underlying goal was to normalize and structure Azure deployments — give customers a framework where, if they follow this one process, their infrastructure will be secure at the end of the day. It’s a public repo on GitHub under Microsoft AI Labs.

What I’m interested in is how we take experimental approaches and make them repeatable. I can envision a structured process for AI experiments where, as a manager, I can make a knowledgeable decision based on outcome — where the value I’m seeking is within reach. Every experiment has a consistent, measured process that can show me the results.

Cloud environments appear very complex, because we’ve moved the infrastructure into the cloud and our breadth of understanding has broadened greatly. But the idea of picking from a library of well-known standards that are certified to be secure — that’s the right direction. Converting chaos into stability. That’s what we do as integrators.

Integration Love Story

When was the love spark — do you have an integration project where you felt: this is it, this is my career?

I’m going to give a shout out to my partner in crime at the time, Brian Loskin. We had never met, didn’t know each other, and I got brought in to do the enterprise service bus for BizTalk at a customer in California. I’m in the office battling with infrastructure — SQL Server deployment, SCSI adapter cards. Brian was in the next room and suddenly shouts around the corner: “Come here, come here. Look at this. I got ESB itineraries working.”

And he’s just loving it. And I thought: I love being with people like this.

For me, my integration love story is the community. The people you meet. And that’s what draws you together. I just got a ping from Brian recently — he went through some surgery. I’ve got friends I’ve made through this community that I would never have met otherwise. It wouldn’t be the connection without integration.

So it wasn’t the technology that made you fall in love with integration. It was the people who work with it.

Exactly.

Fråga till näste gäst

Who should we talk to next, and what question should we ask them?

Somebody who’s relatively new to me in the community but has had a lot of recent impact: Tom Bonoski. He’s a healthcare integrator out of Chicago. He reached out to me recently — after I left Microsoft he leaned in and said “let me try to help you find a job.” It goes right back to the people.

And the challenge question for him — or for anyone you bring in: does integration define you, or do you define integration?

That’s a really good question. And for what it’s worth: for me, it’s been a two-way street. It matches me.

Avslutning

My question back to you: what’s next? What are you going to do with what we discussed today? What’s going to change about how you do things?

Thank you so much, Tom. This has been a really great conversation and we hope to have you again — and hopefully in person in Sweden at some event.

As soon as you have nice weather, just give me a quick call and I’ll come.

The honor is ours. Thank you.