Rethinking the company (Part 2 of 2)
AI can make individual employees more productive. The greater opportunity is to rethink how companies are built, how they operate and how they create value.
ChatGPT can make an individual employee faster. But it does not make the company more intelligent.
So much of the conversation about AI in business still centres on individual productivity. How many employees have access to AI? How often do they use it? How much time are they saving?
These are reasonable questions, but they tell us surprisingly little about whether the business itself is improving.
In a recent global survey, 80% of respondents reported that AI had improved their individual productivity, while only 37% attributed a positive contribution to their organisation's operating earnings. The figures measure different outcomes, but they illustrate an important problem. Companies are becoming better at giving people AI tools than at turning those tools into improvements across the business.
An employee can complete a task faster without the company learning anything from it. A team can use AI every day while continuing to work through the same processes and systems it has relied on for years.
I believe we need to think much more seriously about the company itself. How should work be organised when AI can perform tasks that previously required people? What happens to the knowledge developed through that work? And how might this change the way we build companies and create value for customers?
When individual improvements stay with individuals
Most companies are built around a familiar organisational model. People are employed for their particular skills, responsibilities are divided between departments, and processes determine how work moves through the business.
Over time, employees develop considerable knowledge about customers, products and the problems they encounter. Some of it becomes part of established procedures, but much of it remains in personal documents, conversations and individual working habits.
This has always been a weakness in the way companies operate, and AI makes it particularly visible.
Imagine an employee responsible for preparing customer proposals. They begin using ChatGPT to research products, organise information and prepare drafts. After some experimentation, they develop a method that works well. They know which instructions produce reliable results, which details need checking and how to avoid mistakes that previously took time to correct.
The employee can now complete proposals considerably faster.
But a colleague preparing a similar proposal may have to work everything out independently. The first employee's instructions remain in a personal conversation, and the experience gained through that work never becomes part of the company's established process.
If the employee leaves, much of that knowledge may leave with them.
Making the company better at learning
Every business depends on knowledge about its customers, products, processes and previous decisions. It has accumulated that knowledge through years of work, often at considerable expense.
Yet the information is usually scattered across different systems and people. Even when it has been documented, finding it and knowing whether it is still accurate can be difficult.
A common foundation could bring together relevant documents, business data, established methods and instructions that employees and AI systems can access when carrying out their work.
It would need clear responsibility for keeping information accurate, deciding which knowledge should be retained and managing who is allowed to use it. Without those arrangements, a shared system could simply make unreliable or outdated information easier to spread.
The goal is not to document everything that happens inside the business. Companies need to identify which knowledge matters, preserve what they learn and make it available where it can improve future work.
The company's knowledge provides the context that makes AI useful for its specific operations. AI, in turn, makes that knowledge easier to retrieve, understand and apply.
A new employee can draw on methods developed by experienced colleagues. A team can improve an established process without everyone having to rediscover the same solution. And when the company learns something through its work, it can incorporate that knowledge into the way future tasks are performed.
Over time, the organisation begins to accumulate improvements instead of depending on individuals to remember and reproduce them.
Would we build the company the same way today?
A shared knowledge foundation is important, but the larger transformation involves reconsidering how work is organised in the first place.
Many business processes were designed around the limitations of earlier technologies. Information passes between departments because different people have access to different systems. Employees spend time collecting material, checking routine details and transferring information between applications.
There were good reasons for these arrangements when they were introduced. But they should not automatically determine how companies continue to operate.
If we were building the same company today, with the technology now available, would we organise the work in the same way?
Think about how a customer enquiry moves through a typical business. A request may begin with a salesperson, pass to a technical specialist, require information from operations and eventually involve someone in finance.
Some of those handovers are necessary. Others exist because the information and capabilities needed to complete the work have traditionally been separated between different people and systems.
When information can be retrieved and used more effectively, we have an opportunity to reconsider which steps are still required. Some tasks could be automated, others could be supported by AI, and employees could become involved where their expertise or approval is genuinely needed.
What stands out to me through my work with AI startups is the freedom a new company has to make these decisions before its organisation becomes fixed.
A new venture can build AI into its operations from the beginning. It can decide which work needs to be done by people, which activities can be supported by software and how knowledge should be organised before developing established departments and working habits.
A relatively small team may be able to operate with capabilities that would previously have required a much larger organisation. Research, product development, marketing and administration still need to happen, but providing those capabilities no longer necessarily requires hiring a separate person for every function.
Established businesses face a more difficult challenge.
They already have customers to serve, employees with established responsibilities and systems that the organisation depends on. They have to introduce new ways of working while continuing to operate.
That changes the nature of the transformation. It involves people, established responsibilities and decisions about which parts of the existing business should be preserved or redesigned.
AI agents and the organisation around them
As AI agents become more capable of performing work through connected systems and applications, the way companies organise their knowledge and processes becomes even more important.
An AI agent can be given a task, retrieve information, use software tools and carry out multiple steps on behalf of a person or company. With suitable permissions and safeguards, it can perform work that previously required much more direct involvement from employees.
The agent needs access to reliable company information and clear instructions about how the work should be performed. It must also operate within defined permissions, with human approval where appropriate.
Consider the customer proposal example from earlier.
An AI agent could retrieve the necessary product information, apply approved pricing rules and prepare a draft for an employee to review. If a customer-specific agreement was missing from those rules, the company could correct the information and make the approved change available for future proposals.
Once a process is well understood and can be performed consistently, parts of it may be better handled by conventional software.
There is little reason to repeatedly ask an AI model to reason through a predictable task if the same result can be achieved through a reliable software workflow. That can reduce costs and make the process faster and easier to control.
AI can help businesses explore new methods, deal with less predictable work and develop ways of performing tasks that were previously difficult to automate. As those methods mature, companies can decide which parts should become established software processes.
We can begin to imagine organisations where people, AI agents and conventional software contribute to the same work, each handling the responsibilities they are best suited to perform.
A different way to grow
The relationship between a company's workforce and its capacity to create value has traditionally placed limits on how businesses grow.
More customers often require more service capacity. More transactions create additional administrative work. Expanding into new markets or developing products usually means hiring people with the necessary expertise.
AI changes how much of that work a company can perform with the resources it already has.
For a startup, this could mean reaching the market with a smaller initial team, testing ideas sooner and delaying investments that would previously have been necessary.
For an established business, it might mean serving more customers or developing additional services without increasing its workforce and operating costs at the same rate.
Of course, companies will still need employees with specialist expertise. Growth creates new responsibilities, and the systems themselves require investment and maintenance. But the amount of work an organisation can undertake with a given workforce is becoming less fixed.
This creates opportunities for more capital-efficient businesses, where increasing revenue does not necessarily require a proportional increase in costs.
In many industries, providing a personalised service has traditionally been expensive. Understanding individual requirements, preparing a tailored solution and coordinating delivery all take time. Standardisation has therefore been a sensible way to keep costs manageable.
AI can reduce some of that work, making it more practical to adapt products and services to individual needs.
A service business might be able to prepare more specialised offerings without significantly increasing the time required for each customer. A manufacturer could use its technical information and production systems to support a wider range of product variations.
These opportunities will differ between industries, and standardisation will remain valuable in many situations. But companies now have more possibilities to reconsider how their products and services are designed and delivered.
Building companies that can adapt
The benefits of organisational change extend beyond productivity and financial performance.
A company that depends on particular employees for essential knowledge can become vulnerable when those people leave. A business with poorly understood processes may struggle to maintain quality as it expands.
When knowledge is retained and working methods are shared, those dependencies can be reduced. New employees can draw on what the organisation has already learned, and established processes can be improved without having to rebuild them each time responsibilities change.
Businesses also need to consider how employees will work alongside increasingly capable AI systems.
Some tasks will disappear, while others will change substantially. Companies will need to develop new competencies and reconsider where human expertise makes the greatest contribution.
Professional judgement, creativity and personal relationships will remain important in many areas. In some cases, AI may allow employees to spend more of their time on precisely those parts of their work.
This is one reason I find the European Industry 5.0 perspective relevant. It encourages us to consider what kind of organisation we are building, including the role of people and the ability of the business to withstand change.
For me, that leads to practical questions. Do employees have access to the information they need? Can the company continue operating when an experienced colleague leaves? Are people spending their time on work that benefits from their expertise? And can the business adjust its processes without creating unnecessary complexity?
As AI takes on more consequential work, these questions must also include responsibility and control.
Companies need to decide what AI systems are allowed to do, which information they can access and when human approval is required. Data protection, oversight and accountability belong in the design of those systems from the beginning.
A business that cannot explain or control how its AI systems operate has introduced a different kind of dependency, even if the technology makes its employees more productive.
The company itself becomes a competitive advantage
Two companies can purchase access to exactly the same AI models and get completely different results.
One may use the technology primarily to help employees finish their work faster. The other may have organised its knowledge and processes so that improvements can be retained, shared and developed throughout the business.
Over time, the difference can become substantial.
A company that learns from its own work can improve its methods, develop better products and make more effective use of its resources. Its advantage comes from the experience it has accumulated and the way it puts that experience to work.
Competitors can purchase access to the same AI models. They cannot simply purchase everything a business has learned about its customers or reproduce the way its people, knowledge and systems work together.
This is why I believe we need to reconsider some of our assumptions about building and operating companies. Which competencies do we need? How should responsibilities be distributed? Which established processes still make sense, and what would we do differently if we were starting today?
These are questions I want to explore through my own work. I am interested in building companies where AI becomes part of the organisation, where knowledge can be retained and improved, and where people and technology work together to create better products and services.
If one employee finds a better way of working today, the whole company should be able to use it tomorrow.