Technology Report
|
Резюме
This article is part of Bain’s Technology Report 2026 Everyone expected AI to reshape software organizations by enabling smaller teams, shifting people from doing work to directing agents, flattening organizational structures, and pushing decisions to the edge (see the Bain Brief “How Will AI Change Software Organizations?”). Those changes are underway, but what has been less expected is the speed, scale, and unevenness. Companies have redesigned team structure, redefined roles, and invested faster than during previous operating model shifts. But most companies have addressed only part of the challenge. Without continued effort to reorganize work and how decisions are made, bottlenecks will continue to hold companies back. The organization has already changedThe pace of change is already evident: Bain’s recent survey of software companies found significant shifts in team structures, hiring priorities, and ways of working over the past two years. Fewer engineering organizations report traditional pyramid structures, down from 66% to 29%. Fluency with AI tools is now the leading capability sought out in engineers; cited by 62% of survey respondents, it is three times higher than raw coding skill. Judgment and problem solving also rank highly, reflecting the growing importance of directing agents. The nature of software development is changing rapidly, too. Developers expect to spend a third of their time directing AI agents within two years, up from almost none two years ago (see Figure 1). Job boundaries are blurring, and new hybrid roles are emerging as developers take on testing and deployment, product managers build prototypes and contribute to marketing, and designers write front-end code (see Figure 2). Teams have shrunk to pods of three to five people, with routine execution tasks fading away to be replaced by judgment and management skills often found in more senior roles.
Figure 1
Figure 2
The frontier shows the end stateNative AI companies offer some clues about where this leads since they never had to unwind a legacy organization. Their advantage didn’t start with AI but with good fundamentals: strong test automation, clean requirements, disciplined site reliability, and a culture in which the team that builds something also runs it. Only then did they add agents on top. Companies that skip the basics and reach straight for agentic AI merely exacerbate the problems they already have. But when real change is successful, the numbers are impressive. For example, AI-native marketing organizations generate several times the revenue per employee of traditional software companies by organizing work into small pods, with an embedded engineer and agents doing work that once required entire departments. Of course, knowing the destination isn’t the same as knowing how to reach it. For most companies, the journey starts with a few provocative questions. What shape does the organization settle into? The pyramid is flattening, but the successor is uncertain. The shape depends on what the company is solving for. Some amplify capacity while leaving size and roles largely intact. Some strip out management layers to hasten decision making. Some cut pod size sharply as agents absorb much of the work, leaving a more senior team behind. Others rebuild entirely around small, multidisciplinary teams of generalists. No single answer suits all. Where does the next generation of senior talent come from? The base of the pyramid is disappearing faster than anyone expected. Junior engineers used to make up one-third of teams on average, and now they’re less than a fifth while senior ranks swell. The routine work now performed by AI is the same work on which junior engineers once learned. Companies need more senior judgment than ever, yet they are quietly dismantling the path that produced it. Competitive advantage will go to companies that can compress years of judgment into a fraction of the time by teaching junior engineers how to direct agents. A few already hire this way, recruiting for the ability to talk to customers, solve unfamiliar problems, and build, betting on judgment over pedigree. Who decides now that the chart is flat? More than any other, this is the change that companies have skipped, possibly because this is less about tooling and more about redistributing power. Seventy percent still make decisions centrally, and that applies across different levels of AI maturity (see the Bain Brief “Redesigning Tech Company Operating Models for an AI-Accelerated World”). It’s also the change that matters most because speeding up engineering is wasted if the rest of the organization can’t act on the output. Code ships faster than ever, but the approvals, handoffs, and escalations around it haven’t moved. So, the gains stall before they reach a customer. The fix isn’t another reorganization but rather in distributing decision rights, defining where humans decide, where agents act, and who owns the outcome. One AI-native company lets the engineers building a feature launch its marketing themselves rather than routing it through a central team so that go-to-market moves as fast as the code. Structure changed first. Accountability has to follow, or the flatter organization simply relocates old bottlenecks. What leaders should do nowMost companies have completed the visible redesign: Teams are smaller, roles are changing, and AI is rewriting software development. The next phase will determine how work gets done. Competitive advantage will come less from adopting AI than from building an organization that can fully capitalize on it, with faster decisions, clear accountability, and a talent model that values human judgment over routine execution. Senior executives may be pleased with the progress they’ve achieved, but the harder and more valuable half of the redesign is still a work in progress. More from the report
Read our Technology Report 2026 |