Published: 28 Oct 2026 | Author: James Beresford
Fabric IQ and Work IQ are the most promising match Microsoft has put on screen in years, and most Power BI customers are only watching half the show.
Meet Work IQ. Knows everyone in the building. Reads every email, sits in every meeting, remembers who promised what to whom. Wonderful company. Has never once looked at the accounts.
Meet Fabric IQ. Knows exactly what "gross margin" means at your business, which customers are slipping and what's sitting in the warehouse. Hopeless at small talk.
Put them together and you get something genuinely new. It's an agent that tells your sales director, before Tuesday's client meeting, that the account's margin has slipped four points, two invoices are overdue and the client's ops manager has emailed twice this month about late deliveries. Keep them apart and you've got two very clever singles and no show.
Picture a wholesaler with 50 staff, a finance system, a CRM and a Power BI tenant that's grown up over five years.
At month-end, the finance lead asks Teams which customers to chase this week. The answer comes back ranked by overdue balance, with a note that two of them have gone quiet on email and one has a dispute open with their account manager. Midweek, an agent spots a fast-moving line running low against open orders and messages the operations manager before anyone thinks to open a report. Before a quarterly review, the account manager gets a one-page brief: margin trend, open quotes and the three conversations that matter.
None of that is science fiction. It's Work IQ supplying the people and the context, Fabric IQ supplying the business meaning, and agents doing the joining up.
What's changed is who can afford it. This used to need an enterprise data team and a semantic layer built from scratch. Now the starting point is something most businesses already own – their Power BI semantic models, which Fabric can turn into an ontology instead of rebuilding the business logic all over again.
Honestly – a good first date, but that's it.
With Microsoft 365 Copilot licences, your people can ask Copilot questions it answers from your Power BI semantic models. That works on Pro and Premium Per User as well as Fabric capacity, and row-level security is respected. If your models are well built, it's useful from day one.
But the answer is bounded by what's already in a single model. On Pro, that's whatever your BI person imported, capped at 1 GB and eight refreshes a day. "Customer" in the finance model and "Customer" in the CRM model are complete strangers. And Copilot answers; it doesn't act.
Fabric is where the relationship gets serious.
The Ontology gives your business one definition of a customer, a job or a product, bound across semantic models, Lakehouse tables and real-time data. The graph lets an agent hop across all of it – overdue and complaining and renewing this quarter – instead of answering from one model at a time. Fabric data agents query that data directly, carry your instructions and example questions, and can be published straight into Copilot and Teams. Operations agents go a step further and act on what they see.
Fabric also gives the data that never made it into a report somewhere to live. We saw this recently with a client automation that needed to recognise 15,000 existing projects. It only worked because the project data was sitting in their Lakehouse. The agent you want next year depends on the data you land this year.
There's an SMB-friendly detail on the Work IQ side too. Since the Work IQ API went generally available on 16 June 2026, custom agent usage is billed on consumption through Copilot Credits rather than tied to a Copilot licence per person. Not everyone needs a full seat for an agent to be useful to them.
Every good show has one. This one has three!
First, data processing. Copilot's Power BI answering currently runs its language model processing in the US or EU, so Australian tenants need cross-geography processing switched on. Our understanding is that this is a GPU supply constraint for the Australian region rather than a policy position, so expect it to ease. Until then, it's a decision for your leadership team, not just your admin.
Second, maturity. The Fabric IQ workload is generally available, but the ontology itself is still in preview. Build on it now; just don't hang your month-end close on it yet.
Third, sizing. Graph workloads are hungry, and at least one practitioner throttled an F4 during a demo. Start small, watch the Capacity Metrics app and scale with the use case.
Less than most Power BI customers assume. There's no separate Fabric IQ SKU or add-on fee – it runs on Fabric capacity, and an F2 in Australia East is around $306 a month on pay-as-you-go, before any reservation discount. Add Copilot licences for the people who need them and Copilot Credits for agent use, and the entry ticket for a 50-person business is a fraction of what one analyst costs.
In my opinion, the semantic models you've built on Power BI are the most undervalued asset in your Microsoft estate. Fabric IQ and Work IQ are what finally put them to work across the whole business, rather than one dashboard at a time – and Fabric is where they start earning interest.
Where to from here:
If you're a Power BI-only business wondering what the step to Fabric actually looks like, we'd be glad to walk through it with you, use case first and capacity second.
Will they, won't they? With Fabric, they will :)
Cheers,
James
CEO, Talos
Yes. With Microsoft 365 Copilot licences, Copilot can answer questions from Power BI semantic models on Pro and Premium Per User workspaces as well as Fabric capacity, with row-level security respected. The limit is that each answer is bounded by what's already in a single semantic model.
Yes. Ontologies need a Fabric-enabled capacity, and Fabric data agents need a paid F2 or higher capacity, or Power BI Premium P1 or higher with Fabric enabled. There's no separate Fabric IQ licence – usage is billed through your capacity like any other Fabric workload.
An F2 capacity in Australia East is around $306 a month on pay-as-you-go, before any reservation discount, and Fabric IQ adds no separate SKU or add-on fee. On top of that you'll need Microsoft 365 Copilot licences for the people who use Copilot directly, and Copilot Credits for custom agent usage.
Not yet. Copilot's Power BI answering currently runs its language model processing in the US or EU, so Australian tenants need cross-geography processing enabled in their Fabric tenant settings. Our understanding is that this is a capacity constraint for the Australian region, so it should ease over time.
The Fabric IQ workload is generally available, and Power BI semantic models have been production-ready for years. The ontology item itself is still in preview, so it's worth building on now while keeping business-critical processes on proven components until it reaches general availability.
Get the latest Talos Newsletter delivered directly to your inbox
TECHNOLOGY PARTNERS
Our partners including Microsoft, UiPath, Databricks & Profisee enable us to deliver business outcomes using best of breed technologies and solutions.
Automation & Analytics Technologies for Business
Australian councils face backlogs, staff turnover, and rising community expectations. We help you improve service delivery and operational efficiency through process automation — integrated with TechnologyOne, Civica, and Pathway.
Talos helps organisations build reporting solutions using Jet Analytics
Talos helps organisations build modern data and analytics platforms using Databricks, Microsoft Fabric and Power BI. We work from strategy through implementation, ensuring you get value from the platform without unnecessary complexity.
Work IQ is the intelligence layer over Microsoft 365 – email, meetings, chats and files – that tells Copilot and agents how your people work and who's involved. Fabric IQ is the semantic layer over your business data in Microsoft Fabric, built from Power BI semantic models, ontologies and a graph. Together they let an agent combine the numbers with the conversations around them.