TM1 / IBM Planning Analytics | Modernisation | AI in finance
What I told the Sydney TM1 User Group about modernising TM1, and why it decides whether finance is ready for AI.
By Amendra Pratap, Founder and Managing Director, Octane | October 2026
You can’t ignore AI. If your TM1 models aren’t AI-ready, something else will be, and it won’t only replace TM1. It will replace the people who run it.
That was the message I left with the room at our Sydney TM1 User Group on 21 October, hosted by IBM on George Street. It’s a blunt way to put it, but every finance team using TM1 needs to hear it now, not in two years.
When AI fails in finance, it’s rarely the AI. It’s one of two things. The first is data readiness. AI can’t give you a reliable answer if your data is scattered across spreadsheets with no structure or logic behind it. The second is governance. Finance can’t act on a number nobody can trace, and auditors won’t accept it either.
Here’s the good news. TM1 users start well ahead on the first one. Our dimensions, attributes and aliases already give AI clean, structured data with the business logic built in. That’s a head start most organisations would love to have.
The catch is that a head start only counts if you use it. And that’s where modernisation comes in.
I asked the room how many people were running cubes more than 10 years old. Nearly every hand went up.
That isn’t a criticism. It’s a pattern I see everywhere. We implement TM1 well, then leave it exactly as it was built. Years later someone complains that TM1 feels old school. TM1 isn’t old school. We are, when we stop upgrading and stop rebuilding. And a model built for the business you were a decade ago is not one AI can do much with.
TM1 12 is where all of us are heading, whether we like it or not. Most sites still run TM1 11 as a single Windows process, where a restart means a full reload and one runaway rule can stop everyone. IBM has rebuilt the engine for TM1 12 as separate containers. A crash stays contained, recovery comes from snapshots, and backups run while users keep working.
The honest caveat is that this is a migration, not an upgrade. Architect, Perspectives, Performance Modeler and TM1 Web don’t exist in TM1 12, and TI processes that lean on batch files or local file paths need rework. Setting it up can still take some adjusting. Once it’s running, we’ve found it stable and fast.
There are three ways to run TM1 in our region. You can run it on premises, on IBM Cloud, or as Planning Analytics as a Service on AWS, with Azure following. On premises gives you more control, but more to manage, so it costs more to run. The managed options leave you less to manage and bring new features sooner, in exchange for some control over when changes happen.
So I ask clients one question. Are you a TM1 shop, or a business that uses TM1? Both are valid. But if AI is part of your plans, it matters that new capabilities, especially AI, agents and MCP, land on Planning Analytics as a Service first. Make the choice deliberately, not out of habit.
Don’t be afraid to rebuild
If your model has been propped up for years, breaking it down and rebuilding it is often cheaper and better than patching it again. We all know the moment, a few months after go-live, when we wish we’d added one more dimension.
News Corp Australia is the example I used. Rather than lift and shift its old environment, the team rebuilt the architecture and chart of accounts as NewsPlan on Planning Analytics on Cloud. Queries got 30 to 50 per cent faster, infrastructure, licence and support costs fell by 40 per cent, and the team saves about 260 hours every quarter. Support is the hidden cost here. When you run your own support, it’s hard to see where the time goes.
I finished with a simple sequence. First, assess where you are across model, data, security and who still uses the retiring tools. Second, choose your deployment and your TM1 12 timing. Third, rebuild and automate, from data loads and mapping to exception handling and reporting. Fourth, switch on Workspace, AI forecasting and the assistant.
The order matters. I’ve watched organisations jump straight to AI agents and AI forecasting, get frustrated after three months and then call us. Almost every time, the groundwork hadn’t been done. AI sits at the end of the sequence, not the start.
Kevin Jessop from IBM followed with the Planning Analytics roadmap. Then Steny Sebastian showed what becomes possible once the foundation is right. He asked a TM1 cube questions in plain English, ran a fuel-price scenario and built a dashboard, all without the data leaving Planning Analytics. Every action an agent takes is logged, so auditors can see exactly what it did. That answers the governance problem as well as the data one.
The AI agents Steny demonstrated are now on our website for anyone to explore at agents.octanesolutions.com.au.
Where does your model sit?
If your model was built for a business you no longer are, it’s worth an honest look before AI makes the decision for you. Talk to me on LinkedIn or visit octanesolutions.com.au, and we can work out where to start. And if you were in the room, thank you. The best TM1 ideas usually come from the people sitting next to you.