How should consulting and SI change when scale has become both an asset and a constraint?
Key Takeaways
- AI is not ending consulting or SI. It is shaking the unit by which these services have been sold.
The issue is not that people are no longer needed. The issue is that revenue can no longer be explained as naturally by the number of people and the length of their deployment. - When AI is applied well, delivery efficiency improves. But under the traditional headcount-based business model, that efficiency quickly turns into revenue pressure. And there is no manager who can comfortably absorb that.
Clients will demand shorter timelines, smaller teams, and clearer outcomes. Faster delivery is often discussed as an obvious direction, but for service providers, it becomes as difficult as the transformation they have long advised their clients to pursue. The old way of generating revenue will become harder to defend. - Value-based deals are a necessary direction, but not a universal answer for the entire service market.
Since the rise of digital transformation roughly 15 years ago, value-based deals have repeatedly been discussed in the consulting market. Yet they have not become widely adopted because attribution, measurement, client execution responsibility, and risk allocation remain difficult. VBD may work as an upside option in certain contracts, but it is unlikely to become the base pricing model that supports the whole industry. - The new pricing model is unlikely to be reduced to one single answer.
Human time, expert judgment, reusable knowledge assets, small delivery cells, and operational outcomes may each come to carry their own price. The ability to support new contract structures and new delivery models may itself become a corporate capability. - Large consulting and SI firms do not need to abandon their existing businesses. But they do need a separate operating system for businesses that require different accounting, performance, compensation, and delivery models.
If the new model is managed under the same KPIs as the old model, growth and opportunity capture will become difficult. A more aggressive operating strategy, including potential spin-offs, may be required. This has happened before: in Korea, a major internet company began as an internal venture in the late 1990s; globally, Accenture grew out of a consulting practice that separated from an accounting firm.
A few days ago, I had a chance to speak with a senior colleague in the industry about the future of consulting and SI. As usual these days, the conversation began with AI. Clients use AI. Consultants use AI. Proposals are written with AI. RFPs are also being drafted with AI support. The question is no longer whether we should use AI. Almost everyone does. The real question comes after that.
When AI is used, the productivity of consulting and SI work improves. Research becomes faster. Document drafting becomes faster. Coding and data analysis become faster. But at that exact point, a strange problem appears. The traditional consulting and SI business has explained revenue through the number of people and the duration of their deployment. How many people are assigned for how many months? What grade of resources will be deployed at what utilization level? Under this structure, productivity improvement is clearly good for the client, but it becomes revenue pressure for the provider.
What AI first disrupts is not the reason consulting exists.
What AI first disrupts is the unit by which consulting and SI have been sold.
The old question was relatively simple.
“How many people and how many months are needed to do this work?”
The new question is different.
“How far can this problem be solved, with the smallest possible team, in the shortest possible time?”
That question makes the headcount-based model uncomfortable. Clients already use AI as well. They use it to write RFPs, prepare internal reports, summarize market information, and review vendor proposals. From the client’s perspective, the questions become natural. “Does this still require five people for four months?” “If AI is being used, shouldn’t this be done faster?” “Then why should we pay the same amount as before?”
For the provider, the situation is even more difficult. If AI is not used, competitiveness declines. If AI is used, productivity improves. But once productivity improves, the same work becomes harder to bill in the same way. The better a firm becomes at using AI, the more fragile its old revenue model becomes. This is the uncomfortable paradox now facing the professional services industry.
This is not merely a matter of individual consultant productivity. It is also a management problem. Business unit leaders must manage utilization. Partners must protect revenue targets. Firms must sustain pyramids, compensation structures, and promotion systems. But when the same work can be done with fewer people in less time, it may look like innovation to the client while appearing as revenue contraction in the provider’s income statement. The transformation that consultants have long advised their clients to pursue becomes one of the hardest management problems once it enters their own organization.
There is, of course, an old answer to this problem: the value-based deal. Outcome-based pricing. Success fees. The idea is to create tangible value for the client and receive a portion of that value as compensation. It sounds elegant. In some areas, it is indeed directionally right. Since the rise of digital transformation, this model has been repeatedly discussed in the industry. More recently, in AI projects, terms such as outcome-based and value-based pricing have returned to the conversation.
Yet this model has not become the general pricing model for most service markets. The reason is simple. It is difficult to prove who created the outcome. It is difficult to separate the client’s execution responsibility from the consultant’s contribution. It is difficult to define the baseline. Even when cost savings are achieved, it is hard to determine whether they came from the consulting intervention, market conditions, or internal organizational effort. Revenue growth is even harder to attribute, because sales, product, pricing, channels, brand, and execution capability are all intertwined.
In the end, VBD is a necessary direction, but it is unlikely to become the base pricing model for the entire industry. It can work in certain high-value strategy deals, areas with clearly measurable cost savings, automation projects with quantifiable impact, or engagements tied to long-term operations. But converting every project into a success-fee model is not realistic.
So what is the answer?
To be honest, there is still no clean answer.
And that is precisely why this problem is difficult.
Still, one thing seems clear. The future pricing model will not simply replace time-and-materials with one grand model called outcome-based pricing. It is more likely to become a mix of several pricing units. Short diagnostics and design work may be sold as fixed-price sprints. Repeatable methods, templates, data rules, and industry-specific judgment frameworks may be priced as IP or subscription-based assets. High-end advisory may shift from thick report sales to retainers for judgment and decision support. Implementation work may be sold through small delivery cells that combine a domain lead with AI and data engineers, rather than through large project teams. In some areas, limited outcome-based upside may be added.
In other words, the new pricing model will not arrive as one simple answer. Rather, the pricing unit that was once bundled into one large project may begin to split apart. Human time, expert judgment, reusable knowledge assets, working prototypes, operational outcomes, and decision support may each come to carry a different price.
The important capability here is not only the capability of the individual consultant. The company itself must be able to absorb new contract structures. Legal teams must understand new responsibility models. Finance teams must accept revenue recognition models that differ from traditional man-month billing. Sales teams must negotiate a different pricing logic with client procurement. Delivery organizations must bring fast-moving small teams into a quality management structure. The ability to support contracts and delivery structures that did not exist before may become a capability in itself.
The problem is whether the existing organization can withstand this.
Large consulting firms, SI companies, Big Four firms, MBB firms, and IT service providers all face a similar dilemma. Their organizations are already large. They have many people. They have management systems, promotion systems, utilization management, and revenue targets. But AI-native delivery collides with all of these. Teams must be smaller and faster. They must demonstrate working outcomes rather than only produce documents. Senior people must use AI to raise productivity. But when they do, man-month revenue declines.
This is why many partners try to add AI on top of their existing offerings. On the surface, it looks natural. Add AI to strategy consulting. Add AI to PI. Add AI to SI. Add AI to managed services. But in many cases, this becomes packaging by people who do not deeply understand the technology, the trend, or the current state of the market. They ask AI the wrong questions, then refine the output into consulting language. Clients are beginning to see the difference. At least, they are getting better at seeing it.
There is one scene I find interesting in the field. Clients watch me use AI. They see me spend hours talking to AI, building structures, challenging answers, revising logic, and turning the process into deliverables. But they do not simply see it as “AI wrote it for him.” They see how the questions are asked, what is discarded, how experience is used to sharpen the answers, and where the AI’s response is not trusted. The process itself becomes part of the value.
It is similar to AI-generated content on YouTube. Simply posting output made by AI is different from showing a creative process in which AI is used as part of the work. Consulting is no different. Clients do not pay for AI usage time. They pay for expert judgment, experience, structuring ability, and execution sense amplified through AI.
So what should organizations do?
One classic tool that many people encounter early in business strategy education is the BCG Growth-Share Matrix. The framework, introduced by Boston Consulting Group, looks at a business or product portfolio through two axes: market growth rate and relative market share.
(a) A business with high growth and high share is a Star.
(b) A business with low growth but high share, generating cash, is a Cash Cow.
(c) A business with high growth but low share, requiring selective investment, is a Question Mark.
(d) A business with low growth and low share is a Dog.
Of course, this tool cannot be applied mechanically to the consulting and SI industry as it is. But it is still useful for looking again at the business portfolio after AI. The important question is not whether a business is good or bad. The question is where to invest, where to generate cash, where to reduce exposure, and where to operate differently.
I do not think existing businesses should all be abandoned. Large SI, operations outsourcing, managed services, ERP, and core system implementation will continue to be needed. In many cases, they are Cash Cows. But they should be managed as businesses for automation, standardization, margin protection, and cash generation, rather than as primary growth investment areas.
On the other hand, low-end document production, simple research, generic development outsourcing, and generic PoCs may quickly become Dogs. Clients can already handle much of this internally, and even when they buy it externally, their price expectations will fall rapidly. These areas will become difficult to sell as standalone services.
The Question Marks are GenAI PoCs and agent-building services. Demand is high. But entry barriers are low, and the technology changes too quickly. An agent architecture designed three months ago can already look outdated. If this area is sold as a one-off PoC, it will quickly fall into price competition. It has to be connected to data, business processes, operational accountability, and governance in order to survive.
The Star candidates are elsewhere: AI governance, data architecture, industry-specific AI transformation, AI-native delivery cells, automation for high-risk work, and enterprise AI that combines regulation and security. These areas cannot be solved with simple generative AI usage. They require domain knowledge, data, operations, security, change management, and organizational design.
This is also why the term Forward Deployed Engineer, or FDE, is getting attention. The point is not to copy a specific company’s job title. The point is that we need a very small team that can go deep into the client context, understand the domain, examine the data, build prototypes, and change the actual workflow. An AI-native delivery cell is not a smaller version of the traditional consulting team. It is a different delivery unit in which problem definition and implementation, data and operations, judgment and execution are combined within one team. It may be one person who performs multiple roles, or it may be a two- or three-person cell with complementary capabilities.
From this perspective, what large organizations need is not another innovation slogan. They need two operating systems.
One is the operating system for the existing business.
Utilization, revenue, cost, quality, delivery deadlines, SLA, standardization, and automation matter here. This organization manages existing clients and large contracts. It protects the Cash Cow, defends margins, and reduces unnecessary cost.
The other is the operating system for the AI-native business.
It requires small teams, fast cycles, high capability density, separate pricing logic, separate compensation, separate P&L, and separate quality standards. This organization should not exist to absorb surplus people from the existing business. It should not be a nominal task force created to fill a revenue gap. It should be an independent experimental and business unit that solves client problems in a different way.
A spin-off may be the answer. An internal independent unit may also be the answer. The legal form is not the most important point. The key is that the accounting unit and performance unit must be separated. If the new model is placed under the same revenue targets, same utilization metrics, same hierarchy, and same compensation structure, it will be hard to grow. The existing organization instinctively interprets the new model through the language of the old model. How many people are deployed? How many months? How much revenue? What utilization rate? Once those questions dominate, the AI-native model is pulled back into the old professional services model.
This has happened before. Naver began as an internal venture within Samsung SDS and later grew into an internet company with a very different speed. Andersen Consulting also began as the consulting arm of Arthur Andersen, but required a different growth logic and business structure from the accounting firm. It eventually became independent, went public, and became Accenture. This does not mean every new business must be spun off. But it does show a recurring pattern: businesses that cannot be explained by the management system of the parent often struggle to grow properly inside the parent.
During our conversation, my senior colleague said, “Back to Basic.”
I am not sure I fully understood what he meant by Basic. But after thinking about it several times, I suspect the Basic to which this industry must return is not the old delivery model. It may be a much older set of questions. Why does the client pay? What problem must be solved? How many people are really needed to solve it? How will the result be proven? Who will carry the risk?
Consulting and SI after AI may need to restart from these questions.
Dinosaurs did not become extinct simply because they were large. For a very long time, size was a survival advantage. A massive body meant stability and dominance. There were few environmental constraints, and resources were abundant. So, over nearly 200 million years, evolution favored larger bodies. But when the environment changed, the same body structure became a disadvantage. After the asteroid impact 60 million years ago, a different mode of survival was needed. Smaller animals, animals that could burrow, and lineages that could move quickly became the basis of the next ecosystem.
The consulting and SI industry may be passing through a similar moment.
The existing large bodies will not disappear immediately. Large clients, large systems, and large operating contracts will continue to exist. But size alone will no longer explain the future. What is needed now is a different body structure.
The companies that survive will not be the ones that simply use AI more.
The companies that survive will be the ones that redefine what they charge clients for in a market where AI usage becomes the default.
At this point, I am reminded of the Korean internet industry in the early 2000s. There were users. There was traffic. There were services. But there was no clear revenue model. Everyone was asking, “People are gathering here, but where does the money come from?” Then mobile-phone-based micropayments, cash items in games and online communities, advertising, and subscription models began to combine. Internet services finally started to take the shape of an industry. The core point was not that content suddenly appeared. The unit of pricing was invented for an experience that already existed.
Consulting and SI may now be facing a similar question. Client problems have not disappeared. If anything, they have become more complex. But the way those problems are solved is changing, and the process can no longer be explained only by the number of people and the length of time. Somewhere, perhaps in a small office, a simple shift may emerge: “This is how we should sell it.” It may not be a grand innovation slogan. It may simply be the rediscovery of the unit of pricing.
The question is simple in the end.
How many people should this work require from now on?
What will clients pay for, and how will they pay for it?
Are we still selling human time, or are we selling the unit of problem-solving?
If you are leading a large existing business, perhaps this is where the real question should begin.
Like those hungry people around Seolleung Station in the early days.