Who gets ahead in the enterprise consulting business with AI now in the game?

ON JUN 18, Accenture lost 20 per cent of its market value in a single trading session. TCS fell to a six-year low. Infosys hit levels not seen since 2021.

The commentary called it a guidance cut. The reality was structural: Artificial intelligence was making the traditional IT services model obsolete, and the market was beginning to price that in.

There is a second story inside that same moment, and it is the more interesting one.

While the traditional giants were shedding value, a different kind of company was being built – one that looks like a services firm on the surface but is designed, financially and operationally, to behave like a software company.

The transition from one to the other is one of the largest reorderings in enterprise technology.

Enterprises are stuck in pilot purgatory

Deloitte’s 2026 State of AI in the Enterprise report, surveying more than 3,000 global leaders, found that only 25 per cent of organisations have successfully moved 40 per cent or more of their AI pilots into production. The majority stall somewhere between proof of concept and enterprise roll-out.

The bottleneck is almost never the model. Every enterprise runs on workflows built over decades: custom integrations, undocumented business logic and data scattered across a dozen systems in formats nobody ever standardised.

Before AI can touch a live process, someone has to map those workflows, clean the underlying data, and satisfy security and compliance requirements that vary by industry, geography and regulator.

Then the operational challenges begin: managing model drift, retraining when performance degrades, maintaining audit trails that regulated industries demand.

SEE ALSO

Beyond the backflipping humanoid: where will the robotics returns actually be earned?

AI for life science: Queen’s Gambit or fool’s wager?

Beyond the assembly line – Can Malaysia climb the semiconductor value chain?

Enterprises are not looking for better models. They have access to better models than they can currently deploy. What they want is a vendor willing to be accountable for the outcome end to end.

According to HFS Research, 75 per cent of global enterprises want to renegotiate their IT services contracts. They are fed up with rigid, generic workflows that require armies of consultants and still fail to deliver measurable results.

What they want is an outcome contract: you build it, you own it. That is a categorically different ask from what traditional software vendors or IT services firms are structured to offer.

The model that changes the economics

The companies getting traction share a common architecture. They take on a complex enterprise workflow: invoice processing, regulatory reporting, supply chain forecasting, enterprise resource planning (ERP) migration.

It is delivered not as a team of consultants with a PowerPoint presentation, but as a purpose-built platform combined with a small, specialised human layer of forward-deployed engineers (FDEs) who manage delivery, handle edge cases and feed learnings back into the model.

The platform does the volume. The FDEs own the relationship and the outcome. Revenue can grow faster than delivery headcount.

Traditional managed services firms grow linearly with headcount, gross margins flat at 20 to 35 per cent. The platform model works differently; a system that automates a workflow becomes more efficient with every engagement. The marginal cost of the next client is lower than the one before.

Gross margins expand to 60 to 70 per cent as the platform matures, and grow further as clients add use cases. Because the platform’s data flywheel deepens over time, switching costs become architectural. Retention becomes an engineering problem, not a sales problem.

Palantir has been running this playbook longer than anyone. Its US commercial revenue grew 149 per cent in the second quarter of 2026, total revenue up 93 per cent year on year. The stock has returned more than 2,500 per cent from its 2022 lows.

It spent a decade being called a consulting firm. The market is finally pricing it as the software company it always was.

Big tech bets on the last mile

The surest sign a market is real is when the largest players reorganise around it. In the span of a single month in mid-2026, every major AI company made a structural bet on enterprise deployment.

OpenAI launched a US$4 billion majority-owned subsidiary seeded with embedded engineers; Microsoft committed US$2.5 billion and 6,000 experts to its Frontier Company; and Amazon put US$1 billion into a forward-deployed engineering unit.

The market read the implication immediately. Shares in Accenture, Cognizant and Infosys fell on the day OpenAI announced its Deployment Company. This was not a reaction to a new product; it was a reaction to a new category of competitor.

Companies whose entire pitch had been “buy access to our model” were now building implementation arms and signing outcome contracts, all arriving at the same conclusion: You cannot build durable enterprise revenue without owning how AI gets deployed. The last mile is the margin.

Where the opportunity sits

The richest opportunities share a common profile: workflows that are genuinely complex, data sets that are large and proprietary, compliance requirements that make generic solutions inadequate, and internal AI talent too thin to build the capability in-house.

Gartner puts the global IT services market at US$1.65 trillion in 2026. The companies restructuring around outcome ownership are positioned to capture a disproportionate share of what comes next.

Manufacturing leads the way. Supply chain forecasting and quality control compress planning cycles from weeks to hours. The switching costs that accumulate are not contractual but embedded in the operational data that the platform builds over time.

Financial services follow closely.

Know your customer, anti-money laundering, loan underwriting, claims processing and regulatory reporting are data-rich, compliance-heavy work. Generic solutions fail not because of capability gaps, but because regulators demand a specificity that only vertical specialists can provide.

Healthcare and life sciences add another layer of complexity in clinical data extraction, pharmacovigilance and prior authorisation, where one pharma deployment has already cut three months from a drug approval cycle.

The regulatory depth that makes these verticals hard to enter is the same force that makes specialists nearly impossible to displace once they are in.

ERP and legacy migration rounds out the picture. SAP, Oracle and Salesforce migrations are years-long, expensive and fail at high rates. AI-native platforms are using those initial engagements to earn the trust that expands into broader adoption across the enterprise.

The deepest white space, however, is mid-market. Companies with US$200 million to US$2 billion in revenue are large enough to have rich operational data and real budgets, complex enough to have genuinely painful workflows, and too small to build internal AI capability at the speed the moment demands.

Underserved by hyperscalers chasing enterprise deals and boutique consultants without platform leverage, this is where the category gets built.

Not every company entering this space will build a genuine platform. Three metrics separate the real builders:

• Revenue per FDE rising over time (if headcount scales with revenue, the platform is not automating);
• Time to deploy a new client workflow shrinking quarter on quarter; and
• Net revenue retention above 120 per cent, where deep switching costs show up in expansion, not in renewal negotiations.

The window is open

The disruption of traditional IT services is creating the market for the new model faster than expected.

Every enterprise that once relied on a managed services contract is now asking whether there is a platform alternative that costs less, delivers faster and actually owns the result.

Palantir proved the model works. Amazon, Microsoft and OpenAI validated the demand.

The opportunity belongs to the companies that move first into the complex verticals the hyperscalers cannot serve, build the workflows, secure the data rights and make themselves genuinely hard to replace.

The writer is director, investment, Vertex Ventures South-east Asia and India