Corporate training is worth $400 billion a year. And it still isn’t producing what companies need. In February, The Josh Bersin Company found that 74% of senior leaders say their organizations lack the skills to compete, and fewer than 30% are satisfied with how their workforce is developing. There’s no budget to fix that problem, either. Fosway reports a third consecutive year of flat learning spend and tighter scrutiny of every content line.
Much of that budget has already gone on tools. When we go into an organization to modernize, what we often find is a bloated tech stack with point solutions solving point problems, and none of them fully connected. Every tool holds its own content, its own taxonomy, its own record of who did what. So no leader can answer the two questions that matter — what does this organization know, and what can our people actually do?
We read a number of analyst reports on corporate learning published in the first half of 2026. They don’t agree on everything. How fast to move on AI-native platforms is contested, and we’ll come to that. But there’s broad consensus on the way organizations ought to modernize so we have summarized that in the five steps below.
1. Make skills visible before anything else
According to LinkedIn, 86% of organizations can’t see the skills they currently have. That’s a significant hurdle when skills are still the top driver of learning system change. Today, companies are buying platforms to solve a visibility problem. No platform solves it for you — what makes skills visible is your own data.
Our view: the data that would make skills visible mostly isn’t in a usable state today. It’s spread across an HR system, a performance management tool, and whatever the LMS recorded. Before you map skills, find out what your organization knows and where it’s kept.
2. Move learning into the work
Bersin’s research shows 76% of organizations still run static training: courses built on request, delivered on a schedule, counted on completion. Fewer than 5% have reached what Bersin calls dynamic enablement, where learning, knowledge management and support sit inside the work itself. That small group reports being six times more likely to exceed financial targets, nine times more likely to be rated a great place to build a career, and 28 times more likely to have highly empowered employees.
Our view: what keeps learning outside the work is the way courses are built. A course is more often than not written in advance and fixed at the point of publication. The work however keeps moving and changing, so the content is out of date the moment a process changes, and stays wrong until someone rebuilds or updates that course. That’s workable for compliance, where the material is stable, but it doesn’t work for anything else.
AI changes this. It can draft from the systems that already hold live data — e.g. SOPs, policies, sales records. The material comes from the work, so it can reach people in the work.
Bersin would disagree with us on that point and say that using AI to draft material from live systems doesn’t get you to dynamic enablement. He says it isn’t about using AI to build courses faster, and that it means replacing the SCORM-based LMS with a dynamic content system. He’s right about the first part. Generating the same fixed course in half the time changes nothing. That’s where we part company, and his own research shows why: what makes AI-delivered training work is that it draws on current company and expert knowledge. That knowledge is the input. If it sits across eight systems in eight formats with nobody accountable for it, a new platform has nothing useful to draw on.
3. Enable the people who activate learning
Learning only reaches people if a manager pushes it and the employee can find it. Right now neither is happening. LinkedIn’s workplace learning report puts it starkly: half of employees don’t get the support they need from their manager, and 45% can’t find their way around what’s on offer. A third of L&D teams say they don’t have the resources to help. The role the analysts describe for L&D here is connective tissue — across HR, IT, business leaders, managers and SMEs.
Our view: managers can’t push learning they can’t see, and employees can’t navigate a catalogue spread across eight disconnected systems. Both failures come back to the same thing as step 1. The knowledge isn’t in one place.
4. Re-language the function around business outcomes
Drop the L&D-specific vocabulary: programmes, completions and content. The business doesn’t care about any of it. Training Journal’s L&D Influence Report argues that L&D has to give up its own language and use the words the business already uses: productivity, readiness, quality, retention, risk. Bersin says something similar about how the function operates. When a business unit asks for training, stop taking the order and reporting the completion rate. Work out what the performance problem actually is first and report on that.
We have nothing to add here. It’s absolutely right!
5. Buy AI-native, with the market’s actual state in view
This is where the analysts diverge. Bersin’s recommendation is specific: replace the SCORM-based LMS with a dynamic content system. Fewer than 5% of organizations have done it, and that same 5% is the group reporting that they are six times more likely to exceed financial targets, and 28 times more likely to have highly empowered employees. Fosway, looking at the same market, reports patchy adoption and uneven vendor execution, with delivered AI still behind the hype. It also found fewer than four in ten learning leaders say their platform fits a modern workforce — the lowest sentiment it has recorded.
The 5% who made that change are also the most advanced learning organizations in the Bersin study. So it’s not clear whether replacing the LMS produced their results, or whether the organizations already doing everything else well were the ones ready to replace it. No published data settles that, and it’s an expensive question to get wrong. Bersin himself describes this as a decade-long transition.
Our view: buy AI-native if the case is there, but not before you know what it will run on. An AI-native platform generates from your organization’s knowledge, so it’s only ever as good as what you feed it. That material is usually already sitting in systems you own — recorded customer calls, CRM data, support tickets, SOPs. None of it is learning technology and most of it has never been treated as learning material, but between them those systems hold a record of how the work actually gets done. Find out what’s in there first. A new platform won’t create that record, and it can’t compensate for not having one.
Where this is heading
Vendors are selling a future where the system already knows what your people can do, spots what they’re missing, and gives them what they need at the moment they need it. No course, no assignment. It’s a good future and it isn’t built yet. Most of what’s on the market today as an AI-native platform is more akin to a chatbot sitting on top of a skills map.
It doesn’t much matter whether you think that future is two years out or ten. What you’d do to get ready for it is the same thing you’d do to take step one tomorrow: know what your organization knows and where it lives. Every step above comes back to that.
So: if someone asked you tomorrow to show them where your organization’s knowledge lives, where would you take them?
Organizing corporate knowledge so it can reach people at the point of work is what we do at Sify. If you’re weighing what that looks like across your teams, let’s talk
