How CloudFronts Built a Project Risk Assessment Engine with Databricks Genie

This did not start as a product idea. It started as a recurring frustration inside our own team – the group at CloudFronts responsible for delivery and billing excellence.

Three things kept happening. Early warning signs were buried in long email chains, so risks only surfaced after they had grown. Project managers had no quick way to assess project health, which meant the team reviewed each project by hand every week just to work out where things stood. And because we run many projects simultaneously, it was easy for a real problem to slip through unnoticed.

Here is a real example. One of our clients had a support agreement coming up for renewal. It is the kind of thing everyone assumes someone else is tracking – but it was not flagged until almost the last day. If that client had decided not to renew, leadership would have had no early warning at all. The information was there. It just was not in front of anyone in time.

In another case, a project had two phases and the client had not yet confirmed the scope for the second. That was a clear risk to the final payment. Most of us were aware of it in some abstract way – but the project was not formally flagged until the invoice was already due. A few weeks earlier would have made a real difference.

A project’s health is not in one place. It is spread across four areas – and any one of them, left unmonitored, can become a risk that affects cash flow, client relationships, or delivery quality.

Any one of these on its own does not tell you much. You only see the full picture when you put all four together – and that is exactly what the engine does.

The risk analysis is not a single AI call. It is a deliberate, layered process – lightweight checks first, heavy AI only where it is needed. This keeps it fast and cost-efficient while ensuring accuracy on the things that matter.

Here is how the full architecture fits together – from data sources through to the project manager asking a question in Teams.

Data flows from Dynamics 365 CRM and Outlook → Azure Logic Apps → ADLS Gen2 Medallion layers → Azure Databricks → Genie AI → Microsoft Teams

A project in RED requires immediate PMO attention. RED status is determined by four clear, rule-based parameters – not by a person’s judgment. Any single one is sufficient to trigger the flag.

The output of the Risk Engine is surfaced through Databricks Genie, embedded directly inside Microsoft Teams. Project managers type a question in plain English and get an answer – no tool switching, no separate login.

Genie also works outside its own space. In any existing Teams chat – say, a conversation already happening with a delivery lead or a colleague – a team member can type @Databricks Genie followed by their question, and Genie responds right there in that conversation. No need to open the Genie app or start a new thread.

For example, in a chat with a colleague, a project manager asked: “Show me the invoice details related to the project GA – CSP MSA.” Genie responded in the same chat with the full invoice breakdown – amounts, due dates, and whether each invoice was paid or outstanding. The question did not need to leave the conversation it was already part of.

Four Genie Spaces cover the main areas the team needs. Here are the kinds of questions team members are actually asking:

Genie also supports a tagging feature – team members can tag @Databricks Genie in any existing Teams conversation and ask a question in context. The answer comes back in the same thread without leaving the chat. This makes it a natural part of how the team already communicates.

Here is what it looks like when a project manager asks Genie – right inside Microsoft Teams – to show all projects currently in RED. Genie returns all flagged projects with the specific reason each one was flagged, the risks identified from email and case analysis, and the recommended next actions.

Databricks Genie inside Microsoft Teams – 6 projects flagged RED, each with its risk reason and recommended actions

From the RED list, a project manager can go a level deeper – asking for invoice details, resource utilization, or a full project briefing. Here, Genie is asked to show invoice details for GA – CSP MSA and responds with milestone invoices, outstanding balances, due dates, and payment status – all in one answer.

Worth noting: Genie is not limited to Microsoft Teams. If your organization uses Claude, Genie can be connected and used directly inside Claude as well – the same questions return the same answers, in whichever AI tool your team already works in.

Databricks Genie running inside Claude – the same invoice detail and project insight, available outside of Teams too

The same answers are available on mobile. A project manager does not need to be at a desk to check on a project – they can pull up invoice details, resource breakdowns, or risk summaries from their phone, in plain English.

The same invoice and risk detail – available on mobile, in plain English, without opening any separate tool

The same source of truth answers each team’s own question – without requiring different people to maintain different reports.

Three things need to be in place before this is worth building:

The thresholds CloudFronts uses – health score at or below 5, invoice overdue by more than 21 days, MSA renewal within 30 days, utilization below 50% – were set based on our own delivery patterns and contract types. They are not universal.

A company running large fixed-fee projects might flag an invoice at 14 days. A smaller consultancy running month-to-month retainers might set the MSA window at 60 days. The thresholds are configurable. What matters is that they are agreed across the team and applied consistently – not left to individual judgment on any given week.

Project risk is not a data problem. Most organizations already have the signals – in their CRM, their email threads, their billing records, their support cases. The problem is that nobody has the time to read all of it, every week, across every project, and connect the dots before something becomes urgent.

What the Project Risk Engine does is make that possible. It reads the signals automatically, filters out the noise, applies consistent criteria, and puts the answer in front of the right person – in plain English, inside the tool they already use every day.

We built it to solve our own problem first. It is now part of how we run delivery at CloudFronts. If the problem sounds familiar, the approach is straightforward to replicate – the data is likely already there.

Watch the Webinar

If you want to see the Project Risk Engine in action, we ran a live session walking through the full build – how the data flows from D365 into Databricks, how the four-layer risk analysis works, and how project managers are asking questions in Teams today. The recording covers everything in this article, with a live demonstration.