On July 2, Microsoft announced it is spinning up a new subsidiary, Microsoft Frontier Company, backed by 2.5 billion dollars and roughly 6,000 employees whose only job is to get AI actually working inside customer organisations. Not to build new models. Not to sell more licences. To implement. The Microsoft Frontier Company launch is the clearest admission yet from a major vendor that the hard part of AI is no longer the technology, it is the rollout.
What Microsoft Frontier Company actually is
The new unit, led by Rodrigo Kede Lima, previously head of Microsoft’s Asia business, brings together roughly 6,000 solution architects, deployment engineers, trainers and industry strategists. Engagements run six to twelve months, and Microsoft staff work as what the company calls virtual employees of the client: joining stand-ups, writing code alongside in-house developers and training citizen developers to keep the systems running after they leave. Early customers include the London Stock Exchange Group, Unilever, Land O’Lakes and Novo Nordisk.
The timing is not subtle. Two days earlier, Amazon committed a billion dollars to its own forward-deployed engineering push, a move examined in this related analysis of AWS forward-deployed engineers, and OpenAI and Anthropic both stood up forward-deployed teams earlier this year. The world’s biggest AI vendors are now competing on implementation muscle, not just model quality.
Why the giants are paying for hand-holding
Because the failure mode is universal. Surveys through 2025 and 2026 keep finding the same pattern: companies buy AI tools, run pilots, and stall before production. The blockers are rarely technical, they are process mapping, data readiness, employee resistance and missing governance, the exact issues examined in why AI agent projects stall. When Microsoft prices that gap at 2.5 billion dollars, it is a statement of what the real bottleneck costs.
There is a second signal here for anyone in the consulting business: if the hyperscalers are building implementation armies for enterprises, the same demand exists further down-market, at price points Microsoft is unlikely to serve. The OpenAI partner network made the same bet, as examined in the piece on the rise of AI implementation consultants. The implementation economy is becoming its own industry.
The Frontier playbook, scaled down to a small business
A ten-person firm cannot rent 6,000 engineers, but the structure of a Frontier engagement is worth copying almost line for line.
Embed, don’t advise
Frontier staff sit inside the customer’s daily routines rather than delivering slide decks. A small firm hiring outside AI help can insist that the consultant work inside actual workflows — the inbox, the CRM, the job board — not in a workshop room.
Fix one process end to end
Engagements target specific processes and run until they are in production. The equivalent for a small firm is choosing the single most painful workflow — quoting, scheduling, invoice chasing — and seeing it through before starting a second.
Train the people who stay
The most interesting Frontier detail is training citizen developers so the customer is not dependent forever. Any outside helper should leave behind a team member who can maintain and extend what was built.
Put a timebox on it
Six to twelve months, then handover. Open-ended AI transformation projects are where budgets go to die; dated milestones with a defined end state are where they succeed.
Limitations and what to watch
Several caveats are worth keeping in mind. Microsoft Frontier Company is built for large enterprises: the early customers named at launch — London Stock Exchange Group, Unilever, Land O’Lakes and Novo Nordisk — are all global corporations, and small businesses are unlikely to be direct clients. The announcement is also only days old, so the results cited so far come from Microsoft’s own account rather than independent evaluation. Pricing and engagement terms have not been made fully public. Finally, the unit is designed to work alongside consulting partners such as Accenture, Capgemini, EY, KPMG and PwC rather than replace them, so how the work is divided in practice remains to be seen.
The bottom line
Microsoft just told the market that model access is table stakes and implementation is the product. Small businesses can read that two ways: the tools they already pay for will likely come with much more deployment help, and any AI initiative, internal or hired, can be judged by the Frontier standard, embedded work, one process at a time, skills transferred, deadline attached.
