Deltek Dela AI in Vantagepoint: what it does with your project data

What is Deltek Dela AI in Vantagepoint?#

Deltek Dela AI is the name Deltek uses for the AI features it is adding across its products, including Vantagepoint, its ERP for architecture, engineering and consulting firms. In Vantagepoint, Deltek's announcements have pointed to a conversational assistant, project advisory features and agents that work inside project and CRM workflows. Feature names and scope change between releases.

What your firm actually gets depends on your release, deployment and license. Cloud and on-premises firms may see different features at different times, and some capabilities may sit in add-on modules. The release notes for your version, not a conference demo, are the reliable list.

For a COO or CFO the more useful framing is simple. Dela does not know anything about your firm that Vantagepoint does not already hold. It reads your projects, phases, hours, budgets, invoices and opportunities, and it reflects their quality back to you.

Which Vantagepoint data does each Dela capability read?#

Each Dela capability type reads a different slice of Vantagepoint, which is why data preparation should follow the features you plan to use. The table groups AI capabilities into generic types rather than Deltek product names; confirm which ones your release includes and exactly what each reads.

A practical rule follows from the table. If a project manager would not trust a Vantagepoint report on a project, do not expect Dela to give a better answer on that project, because the assistant reads the same tables the report reads.

What data cleanup should come before turning Dela on?#

The cleanup that matters most before turning Dela on is the routine project and finance hygiene most firms already know they should do. AI features show gaps to more people, faster, so a backlog of stale data quickly becomes a credibility problem for the tool.

The last item on the list is the one firms skip. When a principal asks Dela for backlog and the number differs from the controller's backlog report, someone has to know which definition wins and why, or trust in both erodes.

• 1Inactivate dormant projects and close phases that finished long ago, so questions about active work return active work.
• 2Standardize phase and task codes across project templates, especially on projects converted from Deltek Vision.
• 3Enforce timesheet submission and approval deadlines; advisories built on unapproved hours lag reality.
• 4Set a monthly rhythm for project managers to update estimates to complete.
• 5Merge duplicate firms and contacts in CRM and require a stage and close reason on every opportunity.
• 6Map every user who will use AI features to an employee record and the right security role.
• 7Agree which reports remain the system of record when an AI answer and a report disagree.

Where does your project data go when Dela uses it?#

Where your project data goes when Dela uses it is a contract and architecture question to put to Deltek in writing, not something to infer from marketing pages. AI features in hosted software often call model services that sit outside the core application, and the terms for those calls can differ from the main subscription terms.

Other construction platforms publish specific answers you can use as a benchmark. Procore's AI FAQ, for example, says no customer data is used to train, retrain, fine-tune or improve Microsoft Azure OpenAI or other Microsoft products, while Procore's internal models may use customer inputs and outputs to improve Procore AI accuracy. Ask Deltek for statements at the same level of detail.

Cloud and on-premises firms start from different places#

Cloud and on-premises Vantagepoint firms start from different places because they control their data differently. On-premises firms hold the database and can use the Backup Database utility, which Deltek's help says exists only in the on-premise product. Cloud firms work through Deltek for many changes, and Deltek's Cloud Admin Guide says service requests go through the Deltek Support Center and only from a designated Vantagepoint administrator.

That difference matters when you check what Dela can see and when you want a copy of your own history. In the cloud, list-view exports are scheduled, and Deltek's help says exports of more than 500 records from apps such as Projects, Employees, Firms, Contacts and Activities run on the process server, with your user ID mapped to an employee record. Name the designated administrator before rollout, not during an incident.

Illustrative: a civil engineering firm pilots Dela#

Illustrative: a fictional civil and transportation engineering firm runs Vantagepoint in the cloud after migrating from Vision. The COO wants project managers to stop building weekly status spreadsheets and plans a Dela pilot with a small group of PMs and the controller.

The first test goes badly. Questions about remaining budget return odd answers on converted projects, where Vision-era phases were collapsed into a single phase and old ETCs were never updated. The controller also realizes the PM security role was set up long ago and grants broader cost visibility than the firm now wants.

The firm pauses the pilot to rebuild phase structures on active projects, set a monthly ETC deadline, close dormant projects and tighten roles. The pilot then widens to all PMs and the weekly spreadsheets stop. The same cleanup later makes a metadata-only licensing fit check faster, because project history now reads as a consistent sequence.

Dela runs on your data; licensing it is a separate decision#

Dela runs on your project data to help run the firm, while licensing that history to AI developers is a separate decision with its own rights review and approvals. Turning on a vendor AI feature does not give anyone rights to license your records, and licensing your records does not depend on using Dela.

SourceX approaches Vantagepoint history through the SourceX five-step transaction: Supply, Rights, Preparation, Approval and Delivery. The fit check uses metadata only, such as modules in use, years of history and record families, and nothing is shared at that stage. Payroll, billing rates and HR records typically stay out, and client-confidential material is reviewed project by project.

Under the SourceX Enterprise Data Value Framework, the cleanup that helps Dela also helps a licensing package: data cleanliness and recency raise value, while preparation cost and privacy burden reduce net value. Each approved package is documented in a SourceX Evidence Packet covering provenance, licensing rights, permitted use, the privacy record and release authorization.