On 2 July 2026, Microsoft stood up a new operating business called Microsoft Frontier Company, backed by A US$2.5 billion investment and 6,000 industry and engineering experts who will be embedded inside customer organisations rather than sold as consulting hours. The person running it is Rodrigo Kede Lima, until this appointment the President of Microsoft Asia. Two weeks before the launch he was in Sydney with the EY partnership team, running a working session on how to turn joint clients into what Microsoft is now calling frontier AI organisations.
That combination of facts, the size of the commitment, the delivery model, the leader chosen, and the country he was standing in three weeks before the announcement, is not accidental. Australia sits inside the first wave of the rollout because the local buying pattern (heavily Microsoft-anchored, licensing-led, and increasingly frustrated with production AI outcomes) is the exact market Frontier Company was designed for.
The consequence for an Australian CIO is that the FY27 budget for AI delivery, drafted in most enterprises during May and June, no longer describes the delivery market they are about to procure into. The licensing side has already been reshaped by the DTA's VSA6 agreement. The delivery side is being reshaped in real time by Frontier Company, by Accenture's Microsoft-specific forward deployed engineering practice announced earlier this year, by the parallel AWS Partner-Led Forward Deployed Engineering motion attached to its US$1 billion agentic AI investment, and by the FDE hiring pushes at OpenAI, Databricks, Salesforce and Deloitte in Sydney and Melbourne.
If your delivery model still assumes a systems integrator scoping engagement, a fixed-price statement of work, and a hand-off at go-live, you are budgeting for a world that ended in the first week of July.
What Microsoft Frontier Company actually is
Frontier Company is Microsoft's own version of a practice that started at Palantir and has since been adopted by OpenAI, Anthropic, Databricks, Snowflake and Scale AI. The core idea is that AI systems fail in production for reasons that cannot be scoped in advance, so the engineer who builds the system needs to be inside the customer's environment, working alongside the customer's own teams, owning the outcome rather than the deliverable.
Microsoft describes the model as co-design, co-innovate, deploy and continuously improve. The commercial substance underneath that language is that the same 6,000 people who build the AI system stay to operate it, iterate on it, and remain accountable for measurable business outcomes. Kede Lima will be joined by senior appointments including Kevin Scott (Microsoft's CTO) and Judson Althoff (Chief Commercial Officer) on the leadership bench, which tells you the venture sits between platform engineering and commercial ownership rather than inside either.
Initial named customers include the London Stock Exchange Group, Unilever, Land O'Lakes and Accenture. To scale, Frontier Company will work with the systems integrators most Australian enterprises already use: Accenture, Capgemini, EY, KPMG and PwC. That is not a contradiction, but it is the point of tension the rest of this piece is about.
The Kede Lima Asia signal
Rodrigo Kede Lima ran Microsoft Asia through the period in which Australia became one of Microsoft's fastest growing enterprise AI markets, anchored on the Microsoft 365 Copilot commercial launch, Azure AI Foundry, and the VSA6 whole-of-government agreement effective 1 July 2026. His replacement inside Asia has not been announced at the time of writing, but the fact that Microsoft chose an executive with active Australian client relationships to lead a global forward deployed engineering business is a placement decision as much as a promotion.
The EY-Microsoft Sydney working session in mid June, disclosed on Kede Lima's LinkedIn feed and referenced by both firms internally, focused on the same question Frontier Company was set up to answer. How do we get Australian clients through the gap between a Microsoft 365 Copilot pilot and a system agentic enough to actually change how work gets done. The Frontier Company launch, three weeks later, is the answer.
For an Australian CIO the practical read is that Microsoft account teams will be actively bringing Frontier Company into deal conversations from Q1 FY27, and that the first commercial pitches will lean heavily on named references from Australian clients that already work with EY, Accenture or PwC on Microsoft delivery. If your organisation is one of those references, expect the phone call in the next six weeks.
What a forward deployed engineer is (and what they are not)
The FDE role is not a consulting role, a project delivery role, or a research role. It is a customer-facing production engineer who embeds inside a client organisation, scopes the problem in the client's actual environment, builds the solution against that environment's constraints, and owns time-to-value, adoption, reliability and scalability as measurable outcomes.
That is a different economic contract than the one CIOs have signed with Australian systems integrators for the last twenty years. A traditional SI engagement front-loads scoping into a fixed-price statement of work, transfers implementation risk into the vendor's delivery methodology, and hands off to run-and-maintain when the go-live criteria are met. An FDE engagement inverts that. Scope is discovered through building, not through discovery. Success is measured after deployment, not at go-live. The engineer stays.
The practical implication for Australian delivery contracts is that fixed-price is being replaced by outcome-linked pricing on the AI portion of the stack. Microsoft has not published Frontier Company commercial terms in detail, but the OpenAI and Palantir precedents suggest a mixed model of platform consumption, engineering-time fees, and a component of value share tied to production outcomes. That model does not fit inside a standard SI master services agreement without rewriting the risk allocation clauses.
The Australian systems integrator response
Accenture moved first and moved hardest. In March 2026 it launched a dedicated Microsoft Forward Deployed Engineering practice, positioning itself as the change management, process redesign and industry workflow partner sitting alongside Microsoft's platform engineers. The message to clients was clear: Microsoft brings the model and the engineering, Accenture brings the enterprise-scale delivery, and neither firm is trying to displace the other.
That partnership is durable at the Global 500 level, where change management is genuinely complicated and Accenture's industry depth is the differentiator. It is less clearly durable at the ASX 200 mid-cap tier, where Australian buyers have historically resisted paying twice for delivery layers that appear to overlap. If Microsoft's own engineers are inside the client, and Accenture's are also inside the client, the buyer will ask whose margin they are funding.
EY, KPMG, PwC and Capgemini are all named in Microsoft's launch materials as delivery partners. In the Australian market that named list matters more than in most jurisdictions, because these five firms plus Deloitte hold the majority of enterprise AI implementation revenue between them. Each of them will now be in a competitive framework with Microsoft's own team, coordinated by Microsoft, on delivery work they previously owned.
Deloitte's position is more uncomfortable. Deloitte has invested heavily in its own AI implementation practice under the Zora AI brand and does not appear in the initial Frontier Company partner list. The commercial question that follows is whether Deloitte competes head-on with Frontier Company on Microsoft-anchored delivery, or repositions toward AWS and Google Cloud AI implementation where the platform vendor's own FDE motion is less mature.
The independent Australian SIs (Data#3, Insight Enterprises, Empired, Modis, Nous Group, Kinetic IT and the mid-market Microsoft partner ecosystem) face a sharper version of the same question. If the platform vendor and the Big Four handle the top of the market, and the customer expects an FDE-shaped engagement from anyone selling AI implementation, the local independents need to either build their own FDE capability at speed or specialise into workloads Microsoft's team will not prioritise (regional local government, mid-tier healthcare, mid-tier resources).
The Australian AI talent supply problem is the constraint
Australia produces fewer than 2,000 AI graduates per year at the level required for production engineering, and the projected national shortfall reaches 60,000 AI professionals by 2027. Frontier Company plans to embed 6,000 engineers globally. Even if Australia captures a proportionate share of that headcount (roughly 150 to 200 people), the local talent market cannot absorb the FDE hiring push happening in parallel across Microsoft, AWS, Google, OpenAI, Anthropic, Databricks, Salesforce and every Australian SI at the same time.
That gap is already showing up in salary data. Forward deployed AI engineer roles in Sydney and Melbourne are advertising in a range that starts around AU$220,000 total compensation for mid-level and clears AU$450,000 for senior appointments, before equity or retention bonuses. That is roughly 1.4 to 1.7 times what the same engineering profile earned in 2024. For a CIO planning FY27, the practical read is that even the notional insourced FDE model (hiring your own two or three FDEs and skipping the vendor) has become an eight-figure line item over a three-year horizon at scale.
The tighter the local talent market gets, the more attractive Frontier Company looks on paper. That relationship between local supply and vendor pricing power is something Microsoft has clearly modelled.
What this means for FY27 delivery budgets
If your FY27 plan currently allocates AI implementation spend as a systems integrator line, plus a licence line for Microsoft 365 Copilot and Azure AI Foundry consumption, plus a services line for run-and-maintain, four adjustments deserve serious attention in the next six weeks.
First, the SI line needs a subtle split between (a) change management, process redesign and industry workflow work that the Big Four still do best, and (b) production AI engineering that will increasingly be sourced from the platform vendor's FDE team, either directly (Frontier Company, AWS Partner-Led FDE, Google Cloud AI implementation) or through an FDE-configured SI engagement.
Second, the licence line needs a companion line for outcome-linked Frontier Company or FDE engagement fees. Some of that spend will show up as OPEX consumption, some as a value-share on realised savings or revenue lift, and some as engineering-time fees. None of that fits inside a traditional capitalised implementation project, so the CFO needs to be brought into the conversation early.
Third, the run-and-maintain line changes shape. Because the engineer who built the system stays, the boundary between build and run collapses. Some organisations will experience that as an increase in ongoing engagement cost. Others will see net savings because the incident volume drops. Both effects are real. The FY27 budget should model both scenarios rather than pick one.
Fourth, the internal engineering line needs to be honest. If your organisation intends to build serious internal AI capability, that means hiring FDEs into your own team at competitive comp, which most Australian enterprises are not currently budgeted for. The alternative is to accept that the delivery layer is being outsourced, in which case the internal team's role shifts toward being informed clients of the vendor's engineers, which is a different skillset.
The governance layer becomes the hardest part
Behind the commercial discussion is a governance question that has already gone live. AvePoint's 2026 study found 84 percent of Australian enterprises had rolled back or shut down an AI customer communications agent because of a governance failure, 10 percentage points above the global average. Only 28 percent of Australian organisations report full readiness to investigate an AI or agent related incident. Shadow AI breaches cost Australian organisations an average of AU$968,800 more than standard security incidents, largely because the detection window is longer and the scope of exposure is harder to establish.
The Frontier Company model does not automatically solve any of that. If anything, it changes the accountability structure. When a vendor engineer is inside the customer environment, has production access to sensitive data, and is operating the AI system, the question of who is accountable when something goes wrong sits inside the contract, not inside the customer's own risk register. That is a solvable problem, but only if the contract explicitly assigns identity, access, audit, and incident response responsibilities.
Australian CIOs should assume that the standard Microsoft Master Business Agreement will not be sufficient. A Frontier Company engagement is likely to require a supplementary schedule covering non-human identity governance for the embedded engineer, an incident response protocol distinct from the base Azure service commitment, an audit right over the code and prompts produced inside the engagement, and clear treatment of intellectual property in cases where the engineer builds against customer data.
The APP transparency obligation on automated decision making commences 10 December 2026 under the Privacy and Other Legislation Amendment Act 2024. The OAIC has taken a broad, technology-neutral interpretation of what counts as ADM, which means Frontier Company engagements that produce production AI systems will fall inside the obligation from day one of go-live. That is a compliance timeline every FY27 budget needs to reflect, regardless of who is delivering the AI.
What Wai is watching next
Three signals will determine how quickly this restructures Australian AI delivery over the next two quarters. The first is whether Microsoft publishes a commercial framework for Frontier Company that maps cleanly onto Australian procurement rules, particularly for state and federal buyers already inside the VSA6 framework. Without that, adoption inside the public sector will lag behind the private sector by six to nine months.
The second is whether Accenture, EY, KPMG, PwC and Capgemini formalise sub-agreements with Microsoft in Australia that clarify the delivery boundary. If those sub-agreements exist and are commercially transparent, buyers will procure with confidence. If they do not, buyers will hold their FY27 delivery spend in reserve while the market sorts itself out.
The third is the local FDE labour market. Salary inflation of the current shape is not sustainable across a full FY27 hiring cycle. Either the AI talent pipeline expands quickly through migration policy, university partnerships, and lateral conversion from adjacent engineering roles, or Australian buyers accept a permanent reliance on offshore Frontier Company staffing under Kede Lima's Asia framework. Which of those two outcomes lands will define the shape of Australian enterprise AI delivery for the next five years.
ARC, the platform Wai builds and operates, exists to make the organisations doing this work discoverable inside AI systems as authoritative sources on how they solved these problems. The Frontier Company launch is precisely the kind of market event where getting cited by AI answer engines as an Australian authority on enterprise AI delivery matters more than it did a week ago, because the buyers researching their FY27 delivery decisions are now doing that research through Copilot, ChatGPT, Perplexity and Gemini, not through Gartner PDFs.
FAQ
What is Microsoft Frontier Company?
Microsoft Frontier Company is a new Microsoft operating business, announced on 2 July 2026, backed by US$2.5 billion and 6,000 industry and engineering experts. It is Microsoft's own version of forward deployed engineering, embedding engineers inside customer organisations to design, build and operate AI systems on-site rather than selling a tool and handing off implementation to partners.
Who runs Microsoft Frontier Company?
Rodrigo Kede Lima, previously President of Microsoft Asia, was appointed President of Microsoft Frontier Company at launch. He has worked at Microsoft for more than six years and led the company's Asia business through the enterprise AI adoption cycle. His Sydney engagement with EY in June 2026 is a strong signal that Australia is inside the first wave of the Frontier Company rollout.
What is a forward deployed engineer in enterprise AI?
A forward deployed engineer (FDE) is a customer-facing production engineer who embeds inside a client organisation, scopes the AI problem in the client's actual environment, builds the solution against real constraints, and owns time-to-value, reliability and adoption as measurable outcomes. The role was popularised by Palantir and has been adopted by OpenAI, Anthropic, Databricks, Snowflake, Scale AI and now Microsoft. It is distinct from consulting, project delivery and research.
How does Microsoft Frontier Company affect Accenture, EY, KPMG, PwC and Capgemini in Australia?
All five firms are named as Frontier Company delivery partners. Accenture moved first, launching a dedicated Microsoft Forward Deployed Engineering practice in March 2026 with change management and industry workflow as the differentiator. At Global 500 scale that partnership is durable. At ASX 200 mid-cap scale, Australian buyers are likely to push back on paying twice for overlapping delivery layers, which will force a clearer commercial split between platform-engineering work sourced from Microsoft and enterprise-delivery work sourced from the Big Four.
Should Australian CIOs buy AI implementation from Microsoft Frontier Company or from a systems integrator?
The right answer depends on the workload. For production AI systems tightly bound to Microsoft 365 Copilot, Azure AI Foundry, Dynamics 365 or the broader Microsoft stack, Frontier Company will offer the tightest platform integration and the fastest access to Microsoft's own engineering. For AI programs that require deep industry knowledge, change management, and cross-vendor integration (SAP, Salesforce, ServiceNow, custom data platforms), a Big Four or specialist SI will still be the better lead. Most Australian enterprises will end up buying both, which is why the FY27 budget needs to split the AI delivery line rather than treat it as one number.
How much does a forward deployed engineer cost in Australia?
Forward deployed AI engineer roles in Sydney and Melbourne currently advertise in a range that starts around AU$220,000 total compensation for mid-level and clears AU$450,000 for senior appointments, before equity or retention bonuses. Vendor-supplied Frontier Company or Palantir-style FDE engagements do not publish standard rates, but the observed enterprise commercial pattern is a mix of platform consumption fees, engineering-time fees at roughly 1.5 times normal SI blended rates, and a value-share component tied to production outcomes. Insourcing two to three FDEs into a permanent internal AI team is realistically an eight-figure line over three years for a large Australian enterprise.
What is the AI talent shortage in Australia?
Australia produces fewer than 2,000 AI graduates per year at the level required for production engineering. The projected national shortfall reaches 60,000 AI professionals by 2027. Every hyperscaler (Microsoft, AWS, Google), every frontier lab (OpenAI, Anthropic, Databricks) and every SI is now hiring FDEs into the same local labour pool, which is why FDE salary inflation is running at roughly 40 to 70 percent above 2024 levels for equivalent roles.
How should Australian CIOs budget for AI implementation in FY27?
Four adjustments matter most. Split the SI line between change management and platform engineering. Add a companion line to the licence budget for outcome-linked FDE or Frontier Company fees. Model the collapse of the boundary between build and run in the run-and-maintain line, with scenarios for both cost increase and cost decrease. Decide honestly whether internal AI capability is being built or outsourced, because the two paths require completely different budget shapes. Reflect the 10 December 2026 APP transparency obligation on automated decision making in the compliance and audit line for every production AI system delivered in FY27.
Why do AI pilots fail in Australian enterprises?
AvePoint's 2026 Australian data shows 84 percent of enterprises rolled back or shut down at least one AI customer communications agent because of a governance failure. The dominant failure modes are non-human identity gaps (agents with too much access), thin evaluation and observability at go-live, absent incident response for AI-specific failure modes, and data foundations that were never engineered for the way agentic workflows actually consume them. The Frontier Company thesis (embed the engineer, own the outcome, stay after go-live) is a direct response to that failure pattern, which is why it will resonate commercially in the Australian market even where the pricing is uncomfortable.
Why is Microsoft embedding engineers inside customer teams?
Because the boundary between platform capability and customer outcome moved. When AI systems only work if the model, the data, the workflow and the change management are tuned to a specific customer's environment, the vendor can no longer ship a product and hand off implementation. The FDE model is Microsoft's admission that the pilot-to-production gap is now the binding constraint on enterprise AI adoption, and that the fastest way to close it is to put Microsoft's own engineers inside the customer environment for as long as it takes for the system to work.