The big shifts in how software gets delivered rarely start at systems integrators. Agile didn't. Lean didn't. Six Sigma didn't. For decades that didn't matter much. It's about to.
I've spent the last ten years leading practices at large SIs, most recently one of about 300 people across two continents. The model isn't broken. It's working exactly as designed, and that's the problem.
The kitchen never closes
Ferran Adrià's El Bulli was named the best restaurant in the world five times, and it was only open about six months a year. It got around two million reservation requests a season for roughly 8,000 seats. Most operators would have opened year-round. Adrià shut the dining room for half the year and moved his team to a workshop in Barcelona, where every dish on the next season's menu was born. The restaurant reportedly ran at a loss for years. He paid for invention with revenue he chose not to take.
SIs run the opposite kitchen. The dining room is open twelve months a year, every table is full, and the best cooks are on the line every night. Clients pay to get their hardest problems solved, so the people who can solve them get pulled toward those problems and stay there. Nobody rolls their top architect off a strategic account to spend six weeks learning something new. The client would notice, and the P&L would show it by month end.
The numbers don't leave room
SPI Research's latest benchmark of 509 services firms found billable utilization fell to 66.4% in 2025, the lowest in its 19-year history, with EBITDA around 10% against the 15% SPI considers healthy. And falling utilization doesn't mean your best people have free time. The bench fills with people waiting for the right project while the architects everyone asks for by name stay overbooked.
So internal investment gets cut first. The big Indian integrators put roughly 1% of revenue or less into R&D, and Accenture was around 2% in 2023. Microsoft runs closer to 12%. Even the back office goes without: in Certinia's 2026 survey of 1,000 services leaders, mostly in IT services, 62% still manage projects in spreadsheets. These are firms that sell automation for a living.
SIs spread ideas. They rarely start them.
Look at where delivery innovation actually came from. Extreme Programming, the practice set that kicked off agile, came together in 1996 on a Chrysler payroll project outside Detroit, after Kent Beck was brought in to rescue a system that hadn't printed a single paycheck. Lean came out of Toyota's factories. Six Sigma started with an engineer at Motorola in 1986, and GE made it famous after Jack Welch put it at the center of his strategy.
Every one was born inside a company solving its own problem. The SIs showed up afterward, built practices and certifications, and carried the ideas everywhere else. That's real value. SIs are the distribution network. But they circulate ideas far more often than they originate them. For forty years that was fine, because the inventors had day jobs and weren't competing for the work.
This time the outsiders sell services
AI breaks that deal. The cycle runs in months now, so a firm that's a year behind is working with tools two or three generations old. Worse, the work AI does best is the work SIs sell as leverage: the junior and mid-level hours that make the pyramid profitable.
The firms selling AI transformation are struggling to do it to themselves. In a General Assembly survey of large services firms, 73% of consulting firms had abandoned an AI project because they lacked the skills to finish it, and more than half of respondents said partners were the least prepared group of all.
Meanwhile, the outsiders aren't just inventing this time. They're building services firms. Small AI-native integrators are going after the same work with lean teams and delivery built around agents, competing on speed instead of price. General Catalyst set aside $1.5 billion to build AI-native companies that buy traditional services firms and automate them, including Titan, an IT services roll-up that expects AI to handle up to 30% of client workload. These firms have no pyramid to protect, and their best people spend their time building the system, because the system is the product. That's El Bulli's model, aimed at your clients.
Open the test kitchen
If you run a practice, you can already see the fix: rebuild delivery around agents the way the AI-native firms have. The problem is who builds it. Turning agents into a delivery model that holds up on real client engagements is R&D, and R&D is exactly what the model has no room for. At 10% EBITDA and 1% of revenue in research, nobody's funding a year of your best architects building internal IP instead of billing. So it doesn't get built, or it gets built on nights and weekends and dies the week the next deal closes.
That's the gap W(AI)kforce fills. We've done the R&D so your practice doesn't have to. We deploy named AI agents that join your delivery teams on subscription, each with a defined role and a job description, working real engagements alongside your people: test automation, documentation, code review, estimates, status. Your teams deliver faster on fewer hours, and your practice starts to look like the AI-native firm pitching your clients, without a platform to build or a pyramid to tear down.
The hours you get back don't have to disappear into utilization, either. Pull a few of your best people off the line on purpose and give them the test kitchen they've never had.
Adrià gave up half his revenue to buy time to invent. You don't have to. But someone in your market is already building a kitchen that never needed to close, and they want your best tables.
