Implementing AI without a strategy is like buying a plane without a pilot

Nikola, please introduce yourself in a few words. What is your professional path and how did you come up with the idea to create Invisio?

The idea for Invisio was born partly out of necessity. After the birth of our twins, my free time sharply decreased, and I started actively looking for ways to do more work in less time. I came across a video on YouTube explaining how Google Sheets can be connected to AI to automate a given task, and this was back when ChatGPT was still in version 3.5. At that time, I was working at a product company, and thanks to such automations, I was able to complete my monthly work in just four working days. By the end of 2024, the company went bankrupt, and finding a new position proved to be extremely difficult, and Invisio was born as a logical continuation.

How does Invisio differ from the dozens of companies that today offer AI services and consultations?

In recent months, we have noticed a trend of more companies turning to us after being dissatisfied with their work with other AI consulting agencies and seeking help to solve more complex business problems. In this regard, I suppose that what makes us better than the others is that once we commit to something, we do not leave it until the client is satisfied, and that we never sell hype, but only working solutions for the business.

You often talk about personalized AI solutions rather than universal tools. Why do you think the “one tool for all” approach rarely works in real business?

From the very beginning of the company, we had the opportunity to work with large companies that have many daily interactions woven into complex business processes, and this helped us grow so quickly and to such a level.  Observing how large companies manage their volume and processes, we see that even when universal (SaaS) tools exist, every business has specific needs that the standard solution simply does not fully meet. That's why the need for a personalized AI approach is crucial, especially when working with larger organizations.

What are the most common problems companies come to you with when seeking AI optimization?

We have built a strong reputation in developing specialized solutions for the e-commerce industry, and the majority of inquiries we receive are indeed in this direction.

Currently, our team is working on consolidating all our agents into a unified ERP system, which we plan to provide completely free of charge, so that more brands can access basic ERP functionalities without payment, while the use of the agents in the system will be paid, but on a dynamic basis - according to the frequency of use of the respective tool.

What are the signals that a company is truly ready to begin AI transformation, and not just following the latest technological trend?

The answer is relatively simple, and I share it with every client: well-structured information, easily accessible to every employee, is processed quickly and easily by AI and can be built upon sustainably. Conversely, if conversations with clients exist solely in the mind of one employee, and fuel invoices are scattered in folders on someone’s personal computer, then there is no good foundation on which to build an AI transformation.

Many organizations hesitate whether to hire an AI manager in-house or to use an external partner. What are the advantages and disadvantages of both approaches?

This trend is also observed abroad. We increasingly see open positions for CAIO (Chief AI Officer). In my opinion, smaller companies have a greater real need for a Data or Security Engineering role to ensure proper management and protection of the data flow than for a hired AI manager. The reason is simple: a large part of what is currently offered on the market is hype. The models have already mastered the basic mathematical formulas and systems more than a year ago. From here on, development focuses on optimization, more accurate results, and lower costs. Every major announcement is practically marketing aimed at increasing the value of the producing companies. The truth is that for most business tasks, a complex or expensive AI model is not needed, nor tools built around hype. Properly structured data and well-defined workflows yield much better business results.

For what type of companies does it make sense to build an internal AI team, and when is it more reasonable to rely on external expertise?

In my opinion, a serious internal team is needed for companies that work with extremely sensitive information and want to ensure it does not leave the organization’s perimeter. These companies usually have a strong Security team and pay great attention to Compliance.

We had a case where a large German logistics partner with its own AI department turned to us to solve a complex problem. We live in a time when knowledge is spreading extremely quickly, but it is practically impossible to track all the new solutions that appear daily. Therefore, the symbiosis between external consultants and internal teams is extremely important.

Very often, the internal team of a given company misses something that an external consultant notices almost immediately.

What mistakes are companies most often making when choosing an AI strategy?

They follow the hype and constantly change systems instead of building stable processes on which the models can actually work in favor of their team.

What can happen if an organization invests in an inappropriate solution, the wrong consultant, or technology that does not meet its real needs?

42% of AI projects in the last two years have been stopped for this reason - a poor solution for automating the wrong work process. In such cases, the work of both consultants and internal teams risks being wasted. We often observe companies that come to us with a real and significant problem requiring automation, but upon a more in-depth analysis, it turns out that the costs for AI exceed the price of performing the same task by a person, or the risk of error and damage to the company’s reputation is very high. Many consultants keep this fact quiet to win the project. At Invisio, we believe that the success of the client is a priority, so we warn in advance about all possible risks.

I remember a case where a client told me: “It seems you don’t particularly like AI, you speak quite skeptically about it.”

“I would say that I’m not so much skeptical as realistic.” - I replied.

Have you seen cases where companies have spent significant budgets on AI without receiving real returns? What were the reasons?

There have been instances where a client came to us deeply disappointed with a previous experience with competitors. To the extent that they refused to pay until they saw results. This is completely understandable. The hype around the technology is so great that even I, when I started the company, expected quick, almost magical results. In the course of our work, we have encountered the real limitations of the technology more than once, and this has made us much more careful in the promises we make and more consistent in the warnings we give to our clients from the very beginning.

Which processes are most often subject to successful optimization through artificial intelligence?

We have numerous successful cases in the history of the company, but in about 90% of them, AI performs best in classifying data by several criteria. When the criteria are well defined, and the incoming data is consistent, we observe errors on the order of 1 in 40,000 processed queries.

In which business functions does AI bring the fastest and measurable returns – sales, marketing, customer service, administration, human resources, or other areas?

We have developed solutions for almost all the listed areas, but administration, finance, and accounting, as well as customer service, lead in terms of fast and measurable returns. Marketing, on the other hand, remains the most creative and most difficult to automate area.

How can a company assess whether its AI project is successful and what metrics should it monitor?

We primarily monitor three metrics for every project we develop.

For each automation, we measure the execution time and compare it with the time needed for a person to perform the same task. This way, we understand how much time has been saved. Over time, we have also established other important criteria, such as optimizing costs for AI usage and whether the system requires constant maintenance, dependent on multiple other systems. These factors can significantly increase the cost of a given process, and in some cases even necessitate limiting or stopping it.

Trust is a key factor in such projects. How is trust built between a company and an external AI team when the parties are just starting to work together?

We approach every new project as if it were our own to the extent that sometimes we ask ourselves how we manage to achieve such good results for our clients, while we rarely find time to apply the same approach to our own processes. Most of our clients appreciate the fact that we provide 24/7 support for the systems we build.

How do you overcome the natural concerns of managers regarding data security, confidentiality, and control over processes?

This is a frequently asked question, and we approach very carefully when choosing suppliers to ensure peace of mind for every client. Recently, we had a case where we had to store data for no more than six working days, after which the information was encrypted and remained accessible only to AI upon request, but not to a person. In this way, we guarantee the security of our clients' data.

How long does it usually take for a company to see the first real results from AI optimization?

It depends on the size of the company and the number of operations performed by the respective department. When starting work with a new client, we always strive first to optimize the most frequently performed process in the company. This way, we ensure quick and tangible returns. The so-called (quick wins) are usually visible within the first month and a half.

Is using external AI services an expensive pleasure, and what are the most common myths about the cost of such projects?

I have heard a number of alarming stories about abuses by colleagues in the industry, and probably that’s why people approach every new AI implementation cautiously. What I can share with your audience is the following:

Every AI project is software development done ten times faster than before. Logically, it should also be ten times cheaper, as it is significantly more accessible. This applies to the actual time for development, but it does not mean that communication and accompanying processes in providing the service become cheaper to the same extent.

It is also important to note that when planning a budget for development, it is good to allocate a line for maintenance, as fine-tuning the system also requires time and resources.

What is the profile of the companies that Invisio works with most often?

We most often work with companies in the wholesale, retail, and e-commerce sectors. We have built strong systems that save significant time, and often save money from errors thanks to timely detection and quick informed decision-making.

Are there industries where AI optimization brings particularly high value and why?

Without sounding alarming, I believe that the accounting industry has the potential to be almost entirely automated, with minimal need for human intervention.

Let’s talk about the training you conduct for business leaders. What is its main goal?

“AI for Business Leaders” is a training program we recently launched, because we notice a serious gap between the hype and reality in applying AI in business. Many companies approach fearfully, invest in random directions, and spend significant amounts on tools and promises that do not deliver real value. My approach is that complex engineering solutions are actually the simplest, and we always strive to find a quick and easy way to solve complex client problems.

What are the most common questions you receive from owners and executives during these trainings?

We most often receive questions driven by surprise - the participants in our workshop often react with the words: “Does it really work that easily?”

What knowledge and skills should a modern leader possess to make the right decisions regarding the implementation of AI in their organization?

The most important decision for the modern leader remains the same for years: to build a team of talents whose knowledge and skills will allow the business to develop sustainably with the help of AI. There is no more valuable asset than a qualified expert workforce.

In your opinion, which companies will be the winners of the AI revolution in the next five years?

Those that manage to build a strong team of experts, as I mentioned above, and who have the persistence to keep their database organized, up-to-date, and well-structured.

One of the most discussed questions is whether AI will take jobs. What is your view on the topic?

If we look historically, I do not know of a case where a person can dig faster, more accurately, and more tirelessly than a tractor. In the same way, people whose work is based on repetitive but strictly logical processes are exactly those who, in my opinion, have an urgent need to develop new skills. But as we have already discussed, AI does not have consciousness, and that will remain the biggest advantage of humans. Creative work and critical thinking will remain irreplaceable.

Which professions are most at risk of automation and which will become even more valuable thanks to AI?

I have thought seriously about this issue and do not claim to have a categorical answer, but in my opinion, the accounting profession is among the easiest to automate. If you are an accountant and reading this interview, I would advise you to develop your skills towards consulting clients, rather than just in processing their documents. Otherwise, your business will gradually start to lose ground.

As for professions that will retain or increase their value, I believe that people with serious expertise, long-term experience, and a willingness to take on greater risks will be valued more and more. The reason is that AI is extremely effective in repeating the same process millions of times, but in marketing and sales, it is invariably the more creative and innovative thinkers who win.

What is the role of humans in a future where more and more tasks will be performed by intelligent systems?

Since the birth of our twins, my free time has decreased significantly, and probably that was the turning point that directed me towards automating my daily life, so I could spend more time with the most important people in my life. I believe that AI is here to help us spend more time with the people around us, and that this technology will be one of the last major innovations that will actually bring people closer together, rather than separate them.

How do you see the development of AI in Bulgaria over the next few years, and is the Bulgarian business ready for this transformation?

I consider myself a patriot and believe that our country has serious potential for development in the field of AI at a global level. The successes of INSAIT are probably known to a large part of your readers, but if you have not yet familiarized yourself with them, I highly recommend doing so.

As for the readiness of the Bulgarian business, I believe it is still not fully ready, and that is exactly why we organized the training “AI for Business Leaders”: so that more companies can adopt our methodology for successful AI implementation.

My ultimate goal is to help a sufficient number of businesses so that Bulgaria becomes a recognizable leader on the world map in terms of AI Native companies.

If today an executive director tells you: “I receive ten proposals a week for AI, but I don’t know where to start,” what would be the first advice you would give him?

In our workshop “AI for Business Leaders,” we dedicate a significant part of the time to how companies can recognize unrealistic AI promises. We have our own matrix for evaluating such proposals, which allows us to assess how adequate they are or if they are simply bolstered by promises of a universal solution to many problems. I would be happy to provide access to this matrix to any reader of CareerDailyNews, who contacts me.

And finally – how can companies contact you for consultation, analysis, or implementation of AI solutions, and to which email can they send their inquiries?

The website where you can find more information about our company is invisio.agency, and the email where you can contact me and my team is nik@invisio.agency