The Consultant Is Becoming An Operator

For a lot of people in consulting, the job ends at the recommendations. Even this week I was talking to old colleagues and friends in the industry who still have a large number of projects ending at the ppt deck, before the client finds out if any of it actually works.

Looking back, nobody thought it was a scandal, because it was simply how the industry was built. The people who decided what should happen and the people who lived with what happened were different people, in different companies, on different contracts, and the whole commercial model rested on the join between them. I have been on both sides of that join now, and I think it is closing now faster than ever and possibly for good.

Proof last month it's real

On 24 August, TCS agreed to buy MHP, Porsche's own automotive and industrial consulting unit, for an enterprise value of EUR 320m. Alongside it Porsche committed nearly $1.5bn, over five years to TCS and MHP together, tied to deploying AI across its engineering, manufacturing, operations and customer experience, and to building automotive technology and software-defined mobility platforms.

So Porsche is not buying advice about AI and then finding somebody else to build it. The advisers who already understood the business are moving into the same company that will run the technology, on a five-year commitment. Domain expertise, technology and operating responsibility have been put in one place, with a commercial deal to make it work.

It's a deal underpinned by commercial pressures

Porsche is part of Volkswagen, which is under real pressure to cut costs to compete against Chinese competition, increasing tariffs and the cost of electrification. It has already sold its stakes in Bugatti and Rimac this year and scrapped three subsidiaries, including its battery unit Cellforce and its e-bike business, at a cost of more than 500 jobs. With this in mind, MHP is treated as a disposal.

However, I see it a bit differently, I think this is an organisation that is thinking "we need AI across this business" and concluded the commercial answer to this is bringing the existing teams together, it just so happened that the teams were in different companies and so the deal was created.

Why this is not a one-off deal

As I said in my newsletter last week this is happening across multiple organisations, so far I have read about 7 organisations of completely different types arriving at the same conclusion inside a year.

Amazon put $1bn into a new AWS Forward Deployed Engineering organisation on 30 June, aiming to send 5 or 6 pods of engineers into customers for 45-day stints. Its own announcement is unusually direct about what it is competing with: traditional consulting, it says, assesses, recommends and treats each deployment as a standalone project, whereas AWS structures engagements around business results and states plainly that it does not run on billable hours. Two days later Microsoft announced Microsoft Frontier Company, a $2.5bn business embedding 6,000 industry and engineering experts inside customers, which Judson Althoff describes as going beyond forward-deployed engineering and as the largest outcome-driven engineering organisation in the industry.

OpenAI, which owns the models everybody else is licensing, arrived at the same place from the opposite direction. It stood up a separate services firm, the OpenAI Deployment Company, seeded with $4bn, built out of two acquisitions and now employing forward-deployed engineers in the hundreds, and put a further $150m into a partner programme with BCG, Bain and Accenture. Its chief technology officer, Arnaud Fournier, gives the reason as: there has never been a greater gap between what the models can do and what people use them for.

Then Coforge launched Momentuum AI, an operating unit built entirely around forward-deployed engineers and priced on outcomes. IBM and OpenAI announced forward-deployed units of engineers and consultants working inside client workflows. BCG reorganised the leadership of its North American AI and technology work around a unit it calls purely execution-focused.

Clients are already following this approach

Uber's CTO published the method behind what the company calls Agentic Pods. Around 30 of its most AI-proficient engineers were each paired with a domain expert from a business function and given 10 working days: shadow the expert, prioritise, build alongside the person doing the job, validate with others doing the same work, ship it. In two months they ran 16 pods across 16 functions. Capital allocation across 150 cities went from 15 hours to 30 minutes. Financial reports went from 2 days to 10 minutes. Marketing quality assurance went from 2 weeks to 50 minutes.

These times are not the only bit that stood out, it's the way people work together that he described: once you redesign the workflow around AI you eliminate hand-offs, remove unnecessary approvals and replace legacy tooling, and that the workflow rather than the task is the unit of automation. His closing lesson is something I see day in day out in my agentic projects, the best opportunities are never visible from outside, and that you find them by sitting next to the people doing the work and building with them rather than for them.

Why it might not be true

These labs bring the technical capability but do they have the business transformation vision to ensure what's done fits together in a broader scale? Arun Chandrasekaran at Gartner puts it more bluntly: the labs are capable enough technically, but they do not have the eyes and ears of the board and the C-suite, and they cannot hold a conversation about business transformation at that level. I think this is why the consulting firm doesn't disappear from the story. Someone still needs to be thinking about the big picture, aligning the leaders and deciding where the investments go and this has never been about engineering skills.

Natasha Taylor at BCG points at the same thing from the inside: the barrier is rarely resistance, it is the sheer volume of end-to-end work and the level of executive commitment needed before anything returns. Neither problem is solved by putting an engineer in the building.

So the advisory job has not died. What has changed is that doing it without owning what happens next has stopped being a commercial position that clients are going to accept.

The firms are describing themselves this way now

BCG X is now described by its new leader in North America as 100% execution-focused, carrying a much higher degree of measurability and accountability than the consultant most people pictured 10 years ago, hands-on and driving the change rather than only advising it. He says the definition of a consultant is now broader, which is the argument I have been making all year.

In BCG's AI Radar research in January, 72% of CEOs said they were the primary decision-maker on AI in their organisation, roughly double the year before. When the chief executive owns the decision personally, the deliverable stops being a recommendation they can pass down and becomes an outcome they carry. Even more reason for clients to buy end to end programmes that get to the outcomes.

What crossing that line actually feels like

In the work I do now I am on the other side of the join, owning delivery rather than advising on it, and the change was immediately visible. You find out very quickly which of your own recommendations you actually believed.

Some of them hold up. Others were written by someone who, without ever putting it to themselves in those terms, knew they would not be there when the rubber hit the road. I don't see that as dishonesty, because it is what happens when accountability stops at the presentation. Nothing has improved my judgement about what to recommend as much as having to live with what I recommended!

The question to ask before you sign anything

If you are choosing a partner for AI work, there is one question worth more than the 50 slide credentials pack. Ask to meet the people who will actually build and change the operations.

Then look at how they are organised. If strategy sits in one team, engineering in another, change and adoption in a third, each with its own accountability and its own commercial line, then the traditional consulting hand-off has been recreated under an AI label and you will be managing the join yourself. It will not be described that way in the proposal.

And if you are on my side of the table, the version of the question that stings a little more is the one I would put to any consultant I hire: If you had to stay and run the thing you last recommended, would you still have recommended it?

Sources

All figures as at 2 September 2026.

TCS and MHP. "India's TCS to buy Porsche's IT unit, bags 5-year deal worth $1.46 billion", Reuters, 24 August 2026. The EUR 320m enterprise value and expected close in 3 to 4 months, Porsche's EUR 1.25bn five-year commitment, the Volkswagen cost pressure and the Bugatti, Rimac, Cellforce and e-bike disposals, and Piyush Pandey of Centrum Broking on the land-and-expand comparison and the neutral impact. Also reported by CNBC, 25 August 2026.

AWS. "AWS invests $1 billion to embed AI forward deployed engineers with customers", Francessca Vasquez, 30 June 2026, for the structure, the business-results framing and the billable-hours line. Reuters, 30 June 2026, for the $1bn and the stated goal of 5 to 6 pods on 45-day periods.

Microsoft. "Microsoft Frontier Company: AI engineering that amplifies and protects your intelligence", Judson Althoff, 2 July 2026. The $2.5bn investment, the 6,000 embedded experts and the claim to go beyond forward-deployed engineering.

OpenAI. "OpenAI's High-Stakes, High-Touch Push to Make AI Work for Business", Isabelle Bousquette and Belle Lin, Wall Street Journal, 22 July 2026. The OpenAI Deployment Company, its $4bn seed led by TPG, the Tomoro and Northslope acquisitions, forward-deployed engineers in the hundreds, the $150m partner programme with BCG, Bain and Accenture, and the Fournier and Chandrasekaran quotes. The piece discloses that News Corp, which owns the Journal, has a content-licensing partnership with OpenAI.

Coforge. "Coforge launches Momentuum AI", 30 July 2026.

IBM and OpenAI. IBM Newsroom, 13 August 2026. Andy Baldwin quoted in his IBM title, Global Senior Vice President, IBM Consulting. IBM notes that forward-looking statements are subject to change, so this is intent rather than a delivered result.

Uber. Uber's chief technology officer on Agentic Pods, 7 July 2026, for the pod composition, the 10-day sequence, the 16 pods across 16 functions and the three cycle-time results.

BCG X. Business Insider, Lakshmi Varanasi, 10 August 2026, carrying the Palumbo and Taylor quotes. A firm describing its own operating model, built on interviews rather than on independent reporting of outcomes.

The 72% CEO figure. BCG AI Radar 2026, January 2026, a survey of about 2,360 executives of whom roughly 640 are CEOs. Note that it is a January figure and that it is BCG's own research.