Artificial intelligence has made it easier for companies to pursue digital transformation, yet for many established businesses, the expected gains remain difficult to capture. As AI becomes easier to deploy, competitive advantage increasingly comes down to execution: redesigning commercial processes, reducing operational friction and translating technology into measurable improvements in unit economics, profitability and growth.
Dmitrii Litvinov has led commercial transformation and international business development at Smartcat, WeWork and Rakuten, with experience across enterprise software, flexible workspace and e-commerce. In this interview with FinBold, he discusses why adoption metrics are only a starting point, why large companies often move more slowly, and what changes when AI moves beyond individual use and becomes part of company-wide processes.
Dmitrii, many large companies have adopted AI in some form, but that still does not always translate into sustainable business results. What needs to change for AI to become a real source of competitive advantage and growth?
I’d start with what AI is actually good at right now: absorbing large amounts of information, processing it, analyzing it and turning scattered material into something structured.
For a company, that means building a shared information layer rather than leaving employees to use OpenAI or Anthropic tools individually with whatever document or piece of information they happen to have. You can pay for all those licenses and still miss much of the value if you do not build the infrastructure underneath.
That layer should include business processes, standard operating procedures, company knowledge base, customer and financial data, marketing material, sales records and customer and internal communications. A shared system can also surface contradictions in the company’s own information, for example when different parts of the knowledge base conflict, so people can identify and correct them. People can still get results from using AI individually, but without that shared layer, those results are often less complete and less consistent.
Not as good as it could be, though?
Right. Companies that build this kind of shared layer can get much more value from AI. Those that don’t will fall behind.
But what has to change inside the organization itself?
Management has to decide how broadly people should be able to access company information. Some of it obviously has to stay restricted, but in large companies access is often narrower than it needs to be, especially when compliance and internal controls come into play.
The technical side is getting easier. With tools like Claude Code, many integrations can be built in a matter of hours. The harder part is deciding how much of the company’s internal knowledge, past discussions and senior employees’ experience to open up to a wider group. All of that is an asset the company can put to use.
There’s resistance to that because specialized knowledge can protect a person’s role, relationships and standing, and AI puts pressure on all of that. Then it becomes a question of mindset: whether openness and transparency exist in practice, not just as stated values. Without that shift, AI stays a tool for individual tasks rather than becoming part of the company’s infrastructure.
Do you think businesses will eventually measure digital transformation by business performance rather than by how many technologies they deploy?
Yes, absolutely. But companies will likely keep tracking usage for a while, looking at things such as agent counts and adoption dashboards. Those indicators aren’t meaningless – they can be a useful first step toward measuring business impact.
It’s similar to sales, where activity across a team usually correlates with results, even though the best salesperson might need far fewer touches to close a deal. Usage tells you that something is happening, but you still have to connect that activity to an outcome and track the economics closely enough to know whether AI is actually making a difference. Does AI improve the economics of the task, or would it still be cheaper to have junior staff do it?
You can see the effect more clearly where AI’s strengths map directly onto the work. You can already see it in layoffs among software developers. In some cases, those cuts are not about weak business performance – companies simply need fewer junior developers to handle the more routine parts of the work, while they still need people who can grow into senior and management roles.
Law is another example. It involves huge volumes of documentation and specialized knowledge that used to be highly valued and can now be synthesized much faster. Processing and synthesizing information is exactly what AI does well, and those activities are also a large part of legal work. Different industries will move at very different speeds, so there is no single metric or timeline for this.
At Smartcat, what was harder: implementing new digital tools or changing the way commercial teams drove growth?
Changing teams and processes was harder. At Smartcat and at most other technology companies I’ve worked for, people were already used to moving quickly, and management had generally hired people with the right skills who were a good fit for that kind of environment. Switching systems could be expensive or inconvenient, but the technology itself was rarely the main obstacle.
Tools tend to be algorithmic: you do A, you get B. People are more complicated, and commercial teams also have to deal with customers and competitors outside the company. So figuring out how sales or marketing should work means understanding a much wider set of relationships.
I’ve also never treated tools, team structure and processes as separate problems. They’re all part of the same company. And I never walked into a role with a blank slate. There was always something already working that needed to be improved or reshaped.
Many organizations are moving from digitizing individual functions to redesigning the entire commercial value chain. What difference does that make?
Digitizing an individual function already improves efficiency. That’s what CRM did for sales: it made parts of the process easier. But it becomes much more powerful when people across the company can work within shared infrastructure and access the information relevant to them.
CRM data usually sits with sales, but customer conversations contain product insights and competitive intelligence that many companies still treat as sales information. Feed those insights back to the product team consistently, and they can help improve the product, not just how it is positioned and sold. That’s the real difference between improving one function and redesigning the whole value chain.
Can you make that more concrete? What does it actually improve in the business?
Pricing is one of the clearest examples. You can lose value at the final stage of a sale by giving a discount that wasn’t necessary or losing a customer because the price was too high when a different offer might have worked.
The old approach to segmentation was often static: this account has 10,000 employees or a certain revenue level, so it’s “enterprise” and gets a standard price. But a mid-market customer with a stronger need for one specific feature may actually value the product more than a much larger account does. If you combine industry data, product usage and patterns from similar customers, the sales team can adapt the message, pitch and pricing to the account. That’s how technology can change the economics of an individual deal. The challenge is doing that consistently across different processes and functions throughout the organization.
So why do established companies, despite having more money and data, often adopt technology more slowly than startups?
Decision-making is slower, mainly because there are more approval layers, more established processes, more risk management and more people who can delay a change.
Then there’s legacy technology. Large companies have often invested millions, sometimes billions, in systems that are deeply embedded in daily operations. A bank may know its digital offering lags behind a digital-only neobank, but it can’t simply stop operating and rebuild everything at once. It has to replace systems piece by piece, and because AI moves so quickly, something better may already be available by the time one part of that process is complete.
So you’re always chasing yesterday’s benchmark?
Exactly. Regulation and compliance add more friction, especially for public companies. Those controls matter for customers, employees, shareholders and regulators, but they also make rapid organizational change harder.
AI has reduced the cost of digitizing many business processes. What is now the main constraint on profitable growth: technology, organizational structure or leadership mindset?
If you frame the question that way, the cost of building digital products is falling sharply. People even half-joke that we may eventually see a one-employee unicorn.
If that holds, the challenge shifts to business strategy, entrepreneurship and commercial execution. A strong product can still fail because nobody knows about it, or because it’s positioned badly, sold badly or priced badly. There’s no “it sells itself” button. A small number of companies get close to pure product-led growth, but they’re exceptions.
So AI can’t just hand you a ready-made commercial strategy?
AI can help shape the strategy, suggest opportunities and probably produce a decent first version. But implementation is a different problem. Someone still has to build the company, hire and train people, sell and keep adapting. Maybe one day an agent will be able to build and run a company on its own. For now, sales still involve people, and enterprise B2B usually involves several participants.
Will investors increasingly value companies by their ability to turn AI into measurable operational gains?
Public companies are valued partly on fundamentals, including their ability to generate income over time. If AI cuts costs or increases revenue and that ultimately flows through the P&L into net income, the market will price that in.
What the market struggles to separate is performance driven by real capability from performance driven mostly by timing or luck, such as being in the right market when demand shifts. Investors could lean on proxies, like an “AI-native” index measuring how deeply AI is integrated into the company, but those measures only matter because they tend to correlate with profitability. They are not valuation drivers on their own.
I know some investment funds track AI token consumption when evaluating companies. Is that the kind of proxy you mean?
Yes, that’s a good example. Usage shows that the technology is actually being used, and that tends to correlate with business results, though the metric itself isn’t what creates the value.
You’ve worked on digital transformation both before and during the current AI wave. What has changed most in how you think about building a business?
AI has raised the ceiling on what one employee can do. Earlier, digital transformation was still mostly about tools. Those tools reshaped organizations, but they didn’t expand what one person could do nearly as much as AI does now.
The second change is that technology can become a participant in a business process. You can start giving AI responsibility for part of that process, almost as if it were an employee. That’s a very different way to think about technology.
Technology has always displaced labor in one form or another, so I’m not inclined to catastrophize this wave. What feels new is that AI is beginning to take on nonlinear thinking, communication and other cognitive work that previously required a person. That changes how a manager or entrepreneur thinks about what a role is actually for and how a company should be structured.
Can we say that technology companies outperform traditional businesses because they organize their commercial operations differently?
If you compare two companies in the same industry and one is significantly more advanced in its use of technology, I would generally expect that company to perform better. But how much technology matters depends heavily on the industry. In a commodity business, for example, the market price does most of the work regardless of how sophisticated the commercial process is.
There will always be individual exceptions. But statistically, companies that fail to adapt tend to lose ground, especially in businesses where technology sharply changes the cost and speed of doing the work.
Can large established companies close that gap, or will they always lag?
They can close part of it, but large companies will usually remain behind the technology frontier, and that’s normal. Being behind technologically doesn’t mean being behind in revenue or profit. A large company can still be far ahead in absolute terms. But it still has to keep adapting, even if it doesn’t lead every technological shift. There’s a difference, though, between lagging and standing still. If a company stops adapting, it eventually becomes less competitive.
Read more interviews here.
