AI Is Everywhere in R&D Labs. So Why Are Billions Still Going to Waste?
A new report finds that most companies are using AI to crunch numbers, not to make smarter calls. And that gap is costing them dearly.
Key points
- More than one in three organisations spend between 25 and 40 percent of their R&D budget on projects that never reach the market.
- Nearly half of R&D teams estimate that killing a single project during development or testing costs over one million dollars in wasted investment.
- Most organisations currently use AI for execution tasks, such as data analysis, rather than for decision support at the start of a project.
- Researchers and managers say better intelligence at the early ideation and feasibility stage, before serious money is committed, would deliver the most value.
Picture a company spending years and millions developing a product, only to shelve it before launch. That is not a rare disaster. It is standard practice.
A new whitepaper, flagged by IEEE Spectrum AI, finds that more than a third of organisations burn between a quarter and 40 percent of their entire research and development budget on projects that never reach customers. R&D, for anyone unfamiliar, is the internal work companies do to invent, test and refine new products before they go on sale.
So where does the money go?
Most of it goes to projects that look promising early on, then fail late. Nearly half of teams surveyed put the cost of cancelling a single project during development or testing at over one million dollars. That figure covers the salaries, equipment, testing time and materials spent before someone finally pulls the plug.
Late cancellations are the expensive kind. A project killed after two years of work costs far more than one rejected in the first month.
Why hasn't AI fixed this?
AI has not fixed this because most companies are pointing it at the wrong part of the process. The report finds that organisations typically apply AI to execution tasks, things like crunching datasets, running simulations or building models. These are valuable, but they come after the big decisions have already been made.
The harder question, should we build this at all, tends to get answered the old way: gut instinct, internal politics, and whatever information a team can pull together manually.
That is where the real waste hides.
What would actually help?
Respondents were clear. Better intelligence at the ideation and feasibility stage, the very earliest phase when a project is just an idea being tested against reality, would deliver the greatest value. Catching a bad bet in week two is far cheaper than catching it in year two.
AI tools that help teams assess market fit, competitive risk, or technical feasibility before budgets are allocated could cut that waste sharply. Right now, few organisations use AI that way.
For anyone working in a company that builds physical or digital products, this matters. Wasted R&D spending eventually shows up as slower innovation, higher prices, or fewer new products reaching shelves. Getting AI into the decision layer, not just the data layer, is where the economics shift.
Common questions
Does this apply to small businesses, or only large companies?
The report does not restrict its findings to large corporations. Any organisation that runs a formal research or development process, from a mid-sized manufacturer to a software studio, faces the same dynamic: money committed early, decisions made late.
What should a non-technical manager take from this?
Ask your team where AI is actually being used. If the answer is only in analysis after a project is already running, the decision that started the project probably had no AI support at all, and that is the gap worth closing.
