Hi, I'm Charles Costa, researcher, strategist, and founder of Lexora Labs. Each week I dig into the latest research on AI, knowledge management, and the future of work so you don't have to.
EDITOR'S NOTE:
Microsoft, Amazon, OpenAI, and Anthropic have all made the same bet this year, just at different price tags: billions of dollars and thousands of engineers embedded directly inside client companies to physically make AI adoption work. Microsoft's new "Frontier Company" alone is $2.5 billion and 6,000 people. The reason they're spending like this is the same reason IBM's newest CEO study is getting passed around boardrooms: 85% of employees have AI capability available to them, but only 25% use it regularly.
But a 45-day embedded engagement doesn't answer the harder question this week's research keeps circling: adoption isn't primarily a deployment problem, it's a trust and culture problem, and trust can't be installed by a third party. IBM's own playbook calls for orchestrating human and AI judgment, not just maximizing usage. Forrester found the highest adopters win with use-case-specific tools, not one-size-fits-all rollouts. And a related Harvard Business Review Analytic Services study puts numbers on it: 94% of respondents call well-connected data and workflows critical to AI success, but only 27% say they've actually built that connection.
AI adoption is also a moving target in a way past tools never were, both the technology and the roles built around it keep shifting, which makes "are we learning fast enough" nearly impossible to answer with confidence. Closing that gap takes what no embedded team can install: employees given room to redefine their own roles on their own terms, real governance over what the AI is actually working from, and transparency that earns buy-in instead of demanding it.
THIS WEEK'S READING:
Microsoft commits $2.5 billion and 6,000 employees to new AI implementation unit — CNBC
Microsoft is embedding roughly 6,000 engineers and consultants directly inside Fortune 500 clients to fix stalled AI pilots, a $2.5 billion bet that treats adoption as a staffing problem.
Why It Matters: For AI to actually work, employees need to trust it and shape how it fits their own workflows, that's not something a third-party team can hand off on its way out the door.
AWS is running the identical playbook, just on a much shorter clock.
Amazon launches new $1 billion FDE org, following OpenAI and Anthropic — TechCrunch
AWS is putting $1 billion behind teams of embedded engineers who spend roughly 45 days building production AI systems inside client companies before handing the keys back.
Why It Matters: Amazon's own goal is to leave clients "self-sufficient," but a 45-day sprint says nothing about whether the rest of the organization keeps following the practice once the engineers leave.
IBM's newest research suggests the fix isn't more embedded headcount, it's a different set of internal habits entirely.
2026 CEO Study: 5 plays for AI-first transformation — IBM Institute for Business Value
IBM finds an 85%-capable, 25%-adopted gap, and prescribes five plays: embed AI into how teams collaborate and transfer knowledge, integrate it into real workflows, pick tools by use case instead of a one-size-fits-all rollout, and orchestrate when to lean on human judgment versus AI.
Why It Matters: The plays that actually work treat AI as something teams learn to use together, not a tool handed down from IT.
Forrester's survey of 1,500 AI decision-makers backs that up, with one addition: adoption has to serve the customer, not just internal efficiency.
Accelerate Your AI Voyage — Forrester
High adopters share a customer-led focus, a CEO-driven AI vision, and sustained investment in data and platform foundations, while most enterprises remain stuck three years into GenAI without a clear path to transformation.
Why It Matters: Employees need to see how AI improves their specific role to actually use it, and customers still want a human in the loop for plenty of situations, forcing AI on them isn't a strategy.
The research that ties all of this together comes from a related HBR study, and it puts hard numbers on what's actually broken.
Bridging the Readiness Gap to the Agentic Enterprise — Harvard Business Review Analytic Services
94% of respondents say well-connected data, processes, and applications are critical to AI success, but only 27% say those elements are actually connected at their organization, and just 12% say AI is embedded directly in the flow of work rather than living in standalone tools.
Why It Matters: This is the piece most companies skip: effective AI depends on governed, trustworthy knowledge underneath it, not just document crawling, and getting there takes real upskilling, not a spoon-fed rollout.
Every company in this issue is spending real money to buy its way past the human side of adoption. But the independent research keeps landing on the same missing ingredient, and it's not something any embedded team can install from outside: trust, governance, and room for people to redefine their own roles on their own terms. If you're mid-rollout, the exercise worth running this week isn't picking your next AI tool, it's asking your team what would actually make them trust the one they already have.
Have a take on any of this? Reply directly, I read every response.
Until next Tuesday,
Charles Costa, MLIS