The problem
Intel’s IT team frames its job as supporting the company’s product leadership and its Foundry business during a period of transformation and cost pressure; one section of the report is titled “Bending the Cost Curve” (Intel). Generative AI offered productivity gains, but left to themselves employees and business units would adopt it unevenly, through unapproved tools, with duplicated effort and no way to tell which uses paid off.
The report says Intel is in the early stages of realizing the potential of generative AI to streamline processes, so the problem was twofold: give every employee safe tools, and find out which uses deserve investment.
What they did
Intel IT describes a three-part strategy: provide foundational tools, drive their use across the business, then optimize value by teaching employees to use AI well (Intel).
The foundation is an enterprise generative AI platform with two parts: iForge, the backend engine that connects models and data sources, and iGPT, a no-code frontend that lets any employee build and use AI assistants for specific tasks. Routing use cases through this platform, the report says, helps ensure safe, responsible use of approved, internally built capabilities. Intel runs retrieval-augmented generation on its own Xeon processors and is moving factory deep-learning training to its Gaudi accelerators.
The second piece is governance. An AI dashboard tracks every AI use case across Intel through a life cycle from ideation, proof of concept, pilot and development to production, with cancelled projects tracked as a stage too. Approval weighs cost of implementation, redundancy with other solutions, usage and value, and the information security and privacy teams sit in the approval process.
The third is culture. An AI Everywhere program surveys employees’ AI awareness and then offers training and meetups. The report quotes the chief information officer: “It’s a collaboration tool and no substitute for human experience, judgment, or creativity.” Intel has also deployed 25,000 AI PCs through its regular refresh cycle, and is moving toward agentic AI for customer experience, legal contract review and a general-purpose OneAI bot.
Results
All figures come from Intel’s own report and are estimates, not audited results (Intel).
According to the report, 30,000 monthly users have created more than 7,000 personal assistants and hundreds of generative AI use cases, spanning productivity, cost reduction, cost mitigation and incremental revenue. The AI dashboard lists 436 use cases in total, of which 274 are in production.
The dashboard puts a value on the portfolio. For 2024 it shows estimated business value of USD 305M and estimated hours saved of 321.68K. For 2025, the dashboard estimates USD 376M in business value and 696.03K hours saved; these were forward estimates when the report was published in June 2025. The chief information officer, Motti Finkelstein, also wrote that IT enabled more than USD 4 billion in IT business value in the past year, a figure that covers all of IT’s work rather than AI alone.
What didn’t go to plan
The value figures are labeled estimates, and the report does not explain how business value or hours saved are calculated, whether savings were realized in budgets, or what the platform cost. Hours saved do not translate into cash unless work or headcount changes.
The wider company was in deep restructuring while the platform scaled. In June 2025, Intel told marketing staff that much of their work would move to Accenture, which would use AI to manage parts of Intel’s marketing, with “potential headcount reductions, with only lean teams remaining,” according to a staff notice reported by TechRadar (TechRadar). The outsourcing was part of chief executive Lip-Bu Tan’s plan to cut costs and speed up decisions. Building AI capability in-house while outsourcing an AI-heavy function to a consultancy shows how unsettled the build-or-buy choice remains.
The dashboard also counts projects at proof of concept, pilot and cancelled stages, and the report does not say what was learned from the use cases that did not proceed.
Lessons
- Give employees an approved platform before they find their own. Intel routes use cases through iForge and iGPT to keep them on internally built, approved capabilities.
- Track every use case on one dashboard. Intel’s dashboard shows 274 of 436 use cases in production and puts a value on each stage of the portfolio.
- Put security and privacy in the approval path. Intel’s information security and privacy teams take part in approving AI use cases.
- Let people build their own assistants. No-code tools let 30,000 monthly users create more than 7,000 assistants without a central team building each one.
- Publish how you calculate value. Intel’s USD 305M is an estimate with no stated method, which limits how much a board can rely on it.
Adapt it in your organization
Is this for you?
- Employees across functions already use, or want to use, generative AI tools.
- You can host approved models and connect them to internal data under existing permissions.
- A central team can run a portfolio review with security and privacy at the table.
- You can measure at least one outcome per use case, such as hours or cost.
- Leaders will stop use cases that do not show value.
A 90-day version
Pilot: launch an approved internal assistant builder for two business units, register every use case on a single dashboard with an owner and an expected value, and review the portfolio monthly. Owner: chief information officer. Success measure: at least ten use cases in production with a measured, not estimated, hours-saved figure validated by the business owner.
Metric to track
Measured business value per production use case, and the share of registered use cases that reach production. Intel reports 274 of 436 in production and an estimated USD 305M for 2024. For a government example of measuring assistant time savings, see the UK Government case.
Failure modes to pre-empt
- Reporting estimated value as realized savings.
- Letting unapproved tools spread because the internal platform is slower to use.
- Keeping cancelled use cases on the books without recording why they stopped.
- Counting hours saved without deciding what people will do with the time.
Questions to take to your leadership team
- How many AI use cases do we have today, and who can see all of them?
- Which of our AI value figures are measured, and which are estimates?
- Which functions would we build AI for in-house, and which would we hand to a partner?
Executive FAQ
How many Intel employees use its internal AI platform?
Intel’s 2024-25 IT report says 30,000 monthly users have created more than 7,000 personal assistants on the platform.
How much value does Intel say AI delivers?
Its AI dashboard shows estimated business value of USD 305M for 2024 and a 2025 estimate of USD 376M, by Intel’s own calculation.
About the curator & AI disclosure
Curated by Luciano Ayres
Luciano Ayres is Director of Artificial Intelligence at ArcTouch, with more than 20 years of experience in software and technology leadership and specialized studies in deep learning. Connect with him on LinkedIn.
AI-generated briefing. This case was produced through an automated AI research and writing workflow curated by Luciano Ayres. Sources are linked for independent verification; summaries and analysis may contain errors and do not replace professional due diligence.