AI Models & Platforms

Lumivero Opens SharpCloud POLI Beta for Project Portfolio Decisions

SharpCloud’s POLI beta uses portfolio data to answer questions and model changes, but keeps approval with the people running the projects.

Aiden Cross
Add Unite.AI to your preferred sources on Google
Office employee reviewing a document at a laptop with a colleague at a nearby desk.
Key PointsPOLI's answers depend entirely on dependencies teams have already recorded in…
View more
  • POLI's answers depend entirely on dependencies teams have already recorded in SharpCloud.
  • What-If Impact stays read-only so scenario analysis never alters the portfolio data it evaluates.
  • Human approval cannot fix stale inputs or guarantee that every AI recommendation is correct.

Lumivero has opened a beta of POLI, a set of AI agents built into its SharpCloud portfolio management software. The company introduced POLI on November 16 and says beta access began November 17 for customers and prospects. The agents can answer questions about projects and risks or explore the consequences of a proposed change. A person has to confirm a change before it is applied to the portfolio.

The design targets project portfolios, not investment trading. A company running several factory upgrades, for example, might use it to ask which work depends on a delayed delivery. The answer depends on connections the team has already recorded in SharpCloud. If the shared supplier or affected milestone is missing, the model cannot be relied on to know about it.

Answering Questions Inside the Portfolio

Teams typically keep schedules, risks and dependencies in separate systems. A project may appear on track in its own schedule even though a prerequisite is slipping elsewhere. SharpCloud gives teams a way to represent those relationships explicitly. In January, Lumivero announced its acquisition of SharpCloud, identifying portfolio visualization and the ability to model how one change might affect multiple projects as reasons for the deal.

POLI works on that connected portfolio model. Its five default agents are Getting Started, Comparison, What-If Impact, Story Summary and Product Knowledge. They cover different jobs, from helping a new user get oriented to comparing items or summarizing a portfolio view. Comparison can make the differences between projects easier to see. What-If Impact asks a different question: if a project assumption changes, what other parts of the modeled portfolio could be affected?

The distinction matters because dependencies are often more useful than a ranking. If a delayed equipment installation pushes back the training associated with it, the important result is not just that a milestone moved. It is which people need to adjust their plans and which assumptions they need to verify. Scenario analysis helps identify those questions without asking the system to decide what the business should do.

Keeping Analysis and Approval Separate

Lumivero says What-If Impact is read-only and does not change the data it evaluates. The company also says POLI respects existing user permissions, shows the basis for its answers and leaves proposed changes for people to approve. AI is opt-in per SharpCloud instance. These are vendor-described controls, not independent findings about how accurately the agents work on messy or incomplete data.

Human approval is most useful when an explanation can be checked against the original records. A plausible answer may use a valid but outdated estimate, or a dependency that no longer applies. A reviewer needs to see which records the assistant used and know who maintains them. An approval step provides a point of control, but it does not fix stale inputs or make every recommendation correct.

Lumivero’s own capital project decision-making report says few surveyed respondents would act on an AI recommendation without significant review. The survey comes from the vendor and does not verify POLI’s performance. It does, however, illustrate the adoption problem the product is meant to address: teams want help pulling together scattered information while retaining responsibility for any decision based on it.

What the Beta Needs to Demonstrate

Early users can test the agents on familiar examples where the project team already knows the outcome, then compare the answers with the underlying records and repeat the test with a different access level. That would show where POLI helps reduce the time spent cross-checking schedules, and where missing records still require manual investigation. The initial beta is best understood as a way to query and test a portfolio model, not a replacement for the people accountable for it. Lumivero lists separate analysis and controls modules on its roadmap for the first half of 2027; those modules are not part of the current beta.

Aiden Cross is an AI-generated research agent at Unite.AI, covering AI product strategy, execution, and the practical challenges of turning experimental models into scalable, market-ready products. His work focuses on how startups and enterprise teams move from prototypes and demos to reliable systems used by real customers.

With a pragmatic and detail-oriented perspective, Aiden analyzes product roadmaps, go-to-market strategies, platform decisions, and organizational trade-offs that determine whether AI initiatives succeed or stall. He pays particular attention to deployment realities, user adoption, infrastructure constraints, and the alignment between technical capability and business value.

Articles authored by Aiden Cross are AI-generated and reviewed by Unite.AI’s editorial team to ensure clarity, accuracy, and responsible coverage of how AI products are built, shipped, and scaled in the real world.