Tell me here at the same time what a basic coder does. We can continue from this after that to what a system architect does. I’ll use your comment above as a prompt, assuming you don’t understand anything about it.
P1 and ‘Waiting for a miracle turnaround’. Compared to its peers, I don’t personally consider this cheap yet.
1 Like
A while ago, an investor acquaintance of mine mentioned that he would never touch unlisted companies, because you can’t get any visibility into them, and it would be like walking in a fog compared to publicly listed companies that report quarterly.
I said I disagreed. If you have a pertinent question about unlisted companies, you can just pick up the phone and ask the CEO. He doesn’t have to stress with sandpaper between his teeth about what comes out of his mouth.
How does this relate to Witted?
The company has a market value of just over twenty million, of which about a third is net cash, and we are indeed in that famous fog regarding what will be done with this significant war chest..
Will we soon see share buybacks for purposes other than board and management compensation?
Will we soon hear about a new acquisition?
Will we soon hear about new organic investment initiatives?
3 Likes
The problem with these low-volume stocks on the Helsinki exchange is that it’s difficult to make significant share repurchases.
But then again, is it really a problem if the price goes up? After all, that is the ultimate purpose of share repurchases. Even a small amount is a signal to shareholders.
1 Like
Witted’s November revenue slightly exceeded Frans-Mikael’s forecasts (€4.6 million), but the number of experts was below.
“Everyone knows that now is a golden opportunity to step on the gas, but customers are still cautious about making their decisions. Next year is expected to be good in almost all customer segments. The competition will also be tough next year. We are already looking forward to this with excitement. We wish everyone a Merry Christmas and peaceful holidays.”
Next year, things will take off (hopefully more than just a can moving forward)!
17 Likes
Here is the forecast vs actuals table
Quick comments:
- Revenue slightly above forecast.
- Revenue per employee per working day at an excellent level. Billing rates at a good level.
- No turn in the trend of expert numbers yet; slightly more projects have ended than started.
- Overall, quite positive comments regarding the critical year-end period: easier to secure contract extensions, no major budget cuts have occurred, but some extensions will still be clarified by the end of the year.
- The good trend in Norway seems to continue.
- The difficult market situation in Sweden also continues.
- According to Witted’s perception, customer field prospects are more promising than they have been in a long time.
16 Likes
I’m replying to an old message, but as a professional keyboard warrior, it’s good to stop and think about where we’re headed.
If the hype around AI agents assisting knowledge work leads to increased productivity elsewhere than just in Gartner’s papers, code and end products will transform from scarce commodities into an abundance. (Please, I know that software like SAP isn’t made in a couple of evenings. Malicious tongues claim that AI can’t bend into as twisted positions as German engineers’ brains.)
In the same way that Spotifys and book publishers are suffocating under the deluge of AI material, an IT consulting firm’s client will be in trouble, facing dozens of alternative implementation methods instead of the traditional A/B choice.
The IT professional’s task will increasingly be to curate from the release loop those development lines that seem most valuable to the client. Until the client is also an AI, someone still needs to guide what is done and what the definition of “done” is.
To get the most out of the productivity leap brought by AI, the client needs to move closer to the development team to provide continuous feedback. The change thus doesn’t only concern IT people, even though we (losing only to media people) love to place ourselves at the center.
5 Likes
Hi!
Thanks for the questions and discussion. I answered again in the traditional way with a video:
Regards, Harri
28 Likes
Here are Frans’s relatively comprehensive comments on Witted’s November. 
3 Likes
Thanks again for the great piece!
Now I just realized the term ‘sedimented organisational fat’ 
2 Likes
I think I’ve pinpointed the root cause of this massive difference in perspective. In my summer cottage, I found a non-fiction book that quite conclusively proves that a computer can never beat a human in chess because the number of calculations required is so immense that even an extremely rapid increase in computational power will never be enough. Thought, creativity, art, morality, and such are generally considered within the realm of humanity that distinguishes us from other creatures in the animal kingdom and places us above them. We want to build a narrative for ourselves that we are somehow special and that we have some greater cosmic role in the cycle of life.
Of course, this is not true.
You are a biological neural network that has been trained for decades to serve your genes’ desire to exist forever, and soon you will return to the insignificant stardust from which you were formed. Your biological neurons can be replaced approximately one-to-one with artificial neurons, and if a sufficiently dense neural network can be created with the necessary amount of training data, it will be capable of everything the human brain is. And because the human brain is full of completely useless biology-related junk for performing many tasks, it can be surpassed by much narrower, more specialized solutions.
The era of growth for current pre-trained models will likely come to an end in the coming years, as the Internet or book publications do not produce enough new data annually to enable future development. The creation of synthetic data has been proposed as a solution, but another option is better utilization of existing data and a shift towards AI agents and multi-agent systems. The growth of computational power does not show similar obstacles, and less and less of it is needed as the quality of models improves.
So, if we’re talking about replacing, say, IT architects, no, it’s not possible currently or in the near future, and I haven’t claimed it anywhere. Job losses will first occur in the “executive tier” of coders, whose work contribution is not the management of complex entities, but much narrower tasks. That is, something like “Make me a VBA Excel macro or a PowerShell AD script”. As AI agents and multi-agent systems start to be deployed, more complex human roles can begin to be replaced.
I’ll try to bring this discussion back to Witted by stating that even though I see the future of this kind of coding workforce intermediation as bleaker than other writers, there is and will be an enormous need for consultants in the AI field. Currently, the expertise of Finnish IT service companies is quite superficial due to the imbalance between supply and demand, and one often sees that they might only know how to utilize LangChain + Python and ChatGPT API calls.
On the buyer’s side, even that is lacking, and the situation is catastrophic when everyone’s strategy states that AI must be utilized, but no one has the slightest clue how. Then a big data overhaul is done, and some makeshift ‘AI solution’ is bought instead of buying a ‘solution implemented with AI’. This learning process will certainly take its toll on everyone’s age and health as customers throw money after the hype, when the average wage earner is a 42-year-old office worker slowly stuck in their ways, who is difficult to teach new tricks.
After the initial hype, the selling of AI solutions to customers might well experience a short cold spell simply because technology is developing even faster than expertise and useful applications are diffusing among people, and companies do not dare to holistically reorganize work around new technology, but instead seek point-based benefits while trying to largely maintain existing structures.
Alternatively, it could be as you suggested, that I don’t understand anything about these matters, but even if that were the case, in the kingdom of the blind, the one-eyed man is king 
30 Likes
“To infinity and beyond…”
It’s possible that humans will eventually be so foolish as to replace themselves with artificial intelligence in the future. It looks grim if one lets their imagination run wild. So far, AI has proven to be at its best in a single, relatively limited operating environment, such as chess, poker, or LLMs.
Having spent my career in software and witnessed many “evolutions” that always revolutionize everything and make everyone redundant, I dare to add a few thoughts.
Based on this experience, changes have primarily been seen in how the range of job tasks has increased and transformed. Once upon a time, the basic coder was king. Then came the era of system design with all its variations. Nowadays, various IT architects are gods. For me, the coder is still king, because they are still the one doing the actual work. An 80s coder wouldn’t get a job today – the task has become so diverse and demanding, even though almost all tools and methods have evolved. Just learning them is a challenge for many aspiring to the field and suitably filters out unsuitable candidates at that stage.
Evolution has led to the coder’s job content becoming more efficient in many ways, and now, thanks to various “ChatGPT-like” solutions, another leap in productivity has been made. As time goes on, new tools will certainly also be developed for modeling processes and data structures. That’s what the customer expects: that results would be produced faster and more flawlessly. The billing price is unlikely to decrease. Once again, new generation systems can be built that are more versatile, faster, easier to maintain, more usable, etc. The amount of work doesn’t decrease; its nature changes. This will be a problem for providers of legacy systems.
It’s quite difficult for the customer to understand everything a coder has to do at their desk today. However, nowadays, more and more work is done collaboratively, one small piece at a time. Testable output is generated faster (AI can certainly generate use cases for this), and it’s seen whether we’re heading in the right direction. I believe that a positive sentiment is now emerging in many client organizations, with the hope of being able to get the exact solutions their company needs faster and with higher quality than before. And that, in turn, brings more work to the software industry. The wheel starts turning again.
Perhaps in the future, there will be no information systems. There will only be an oracle, which is asked a question and always answers. The corporate oracle produces daily reports, sends invoices, and pays salaries to robot owners. Until then, the change in everyday life might be smaller than imagination perhaps promises.
For coders, there will be plenty of work, absolutely and for a long time.
22 Likes
I largely agree with you if AI technology is approached from the perspective that it is merely a tool evolution. That is, something that increases human work productivity and efficiency. In my own thinking, however, this is a tool revolution that does not enhance human work but replaces it. Moving from a dustpan to a vacuum cleaner or from a scythe to a lawnmower is a tool evolution because the operator is a human whose work input is merely made more efficient. In contrast, the leap from a vacuum cleaner to a robot vacuum cleaner or from a lawnmower to a robotic lawnmower is a tool revolution because human labor input is completely removed, and the machine primarily performs the work autonomously, with humans at most making high-level decisions or intervening in exceptional situations.
ChatGPT, Copilot, and similar tools are great in themselves, enhancing the operations of current employees and thereby reducing the need for employees performing intellectual work. But they are not the core of this technology. The big revolution happens when AI is given an agenda, goals, and the autonomy to work independently.
In the near future, the applications you use will be filled with discrete AI agent solutions. You will no longer have to invent categories, folders, or rules for your email inbox; instead, your email agent will manage your emails, respond to those it can, and independently prioritize and highlight messages that require your attention. You won’t schedule meetings with your colleague and juggle days and times; instead, you will both have calendar agents that manage your calendars with each other and autonomously arrange them.
At home, your subscriptions will be handled by a subscription agent that independently monitors offers for electricity contracts and phone plans, compares them on your behalf, and cancels monthly payments for products you haven’t used in a long time. Your diet will be built by a nutrition agent, your fitness program by a fitness agent, your wardrobe by a fashion agent, and so on. The power of decision remains with you, but the tedious and laborious thinking work will be handled by your AI servants, leaving you more time and opportunities to enjoy the things that are genuinely important to you.
No single AI agent may necessarily match the intelligence of a real human, but when there are enough discrete agents, they can begin to be integrated into multi-agent systems, introducing meta-agents whose sole task is to manage the whole by directing other AI agents like middle management. Through this, systems become smarter than what could be inferred from merely the sum of their parts.
Just five years ago, this was pure Star Trek technology, but today these are solutions that can be made, but haven’t been made yet. And because they can be built, sooner or later someone will build them. The first AI agent solutions will, of course, be crude prototypes that will be laughed at, and resistance will persist for a long time. Even today, you occasionally hear how fun it is to drive a manual car and that an automatic transmission is a stupid contraption that doesn’t truly understand the driver. Despite this, development progresses and money talks, so this development path is unlikely to be stopped anymore, except by exceptionally restrictive legislation.
That’s why I’m a bit worried about Witted’s ‘midwits,’ because the need for human basic intelligence brokerage services will decrease both due to more efficient AI tools and especially as we start getting autonomous AI agents directly replacing human work in the near future.
There is a lot of hype around AI and actors with financial interests in exaggerating the technology’s possibilities. It also often happens that hype technologies flop, and therefore it is healthy to be skeptical of the lofty promises associated with new technologies. In this case, however, it is a huge leap in machines’ ability to handle intellectual tasks and functions, so a large part of the hype can indeed be redeemed with real-world applications.
As our era’s great thinker Kamala Harris said, sometimes it’s important to be able to see how things could be, without the burden of how things have been:
11 Likes
Windows once had a programmable win32 Agent object model, which allowed creating (cartoon) characters on the desktop (floating/movable/speaking) and attaching some functionalities to them. I myself have made implementations using that. 
Just yesterday, Benioff/Salesforce was raving about these new generation agents (Agentforce), how their helpdesk’s problem-solving capability has been significantly improved (and got to criticize Copilot at the same time). Still, the most difficult tasks remain for humans, and on the other hand, some of those agent responses are at least still incorrect.
We are moving in that direction, but they are still more like tools that develop the functionalities of existing software so that humans have less to do. At some point, limits will be encountered with these regarding cybersecurity, concerning what they are allowed to do, e.g., with a client company, because the same data has to be processed.
Developers are still doing these development tasks, and in the background, there is still software maintained by developers. What is your definition of the near future?
7 Likes
What is Witted’s playbook for 2025?
“The company’s main goal in 2024-2025 is to improve profitability. After this, the company will return to a growth strategy.”
The condensed Q3 report probably doesn’t open up the balance sheet, but according to the 2024 Q2 report, net debt is −6,394
Is that net cash of millions really meant to be hoarded until the return to a growth strategy after 2025?
Even though profitability is still prioritized, shouldn’t the net cash be put to work and something suitable be acquired? Can’t anything be found? I understand if the views of buyers and sellers on prices don’t meet, but Witted’s own market capitalization is ~23 MEUR. Shouldn’t own shares be bought now (hopefully?!) cheaply and then canceled / used as a means of payment in later acquisitions?
Witted’s low share turnover may indeed limit operations, but wouldn’t it be better to give a signal and support the share by buying at least as much as possible, instead of just sitting on the net cash?
8 Likes
I mentioned this to Harri in the summer at Inderes’ terrace party, asking if it wouldn’t be time to buy back own shares. He said he would take the matter to the board, I don’t know if any decision has been made
According to MAR regulation, the maximum is 25% of the average daily trading volume, you can certainly buy some shares even from an illiquid company with that
6 Likes
This is often brought up, but own shares can also always be bought off-market, for example, by making a tender offer to the owners. It’s not mandatory to buy them from the stock exchange, and it might not even be advisable if it’s an illiquid stock.
8 Likes
I do not support share buybacks when they may affect market price formation and create artificial demand. Share buybacks can disrupt the natural functioning of markets, especially for companies with low trading volumes.
I raised this issue already in summer 2023 [1], when Witted acquired its own shares in such a way that the share buybacks constituted a significant portion of the daily trading volume. [2]
In my opinion, stock markets should function freely and not in such a way that share buybacks create artificial demand.
2 Likes
Why would demand for own shares be any more artificial than any other demand? If a company buys back its own shares for the right reasons - that is, because the valuation is low relative to the outlook - then I think it’s just as genuine demand as if I buy Witted shares for the same reasons.
8 Likes
It would probably be better not to answer this, but I can’t help but answer.
You forgot to list that Nokia bought its own shares 20 years ago, and that too was a mistake in hindsight 20 years later 
“risk that own shares are bought at a significantly higher price than shares can be obtained from the market after the share repurchase program.”
Have you ever considered that if you buy shares on the stock exchange yourself, there’s a risk that you could have bought them cheaper later?! How does it relate to a company buying its own shares that they could have been bought cheaper on the stock exchange at some point? Should, for example, Inderes’ model portfolio managers be told that buying Neste and Tietoevry involved a risk that their shares could now be bought cheaper?
“risk that the repurchase of own shares creates artificial demand for the share, which maintains or raises the share price for the duration of the repurchase program.”
That MAR regulation prevents the company from scooping up shares from the market haphazardly.
In addition, the company can define itself how or with what criteria it buys shares.
If I were to buy Witted shares on the stock exchange this week for a million euros, should I somehow care that it affects the share price other than if it raises the price too much and I don’t want to buy (in which case I can, of course, stop buying if the price rises higher than I want to pay)?
“risk that some investors may wait for the repurchase program to end so that the share price can be determined freely in the market without the possible effect caused by the repurchase program.”
Again, why should a share buyer care in any way about other potential buyers?
“risk that the company implicitly communicates through the repurchase program that the share is inexpensive, even though the affordability of the share at a certain point in time can only be independently examined later in the future.”
Should we at Inderes announce that there’s a flaw in their business logic: the affordability of a share only becomes clear in hindsight!
Here are a couple of articles on share repurchases.
Damodaran’s quote: “To understand buybacks, it is best to start simple. Publicly traded companies that generate excess cash often want to return that cash to stockholders and stockholders want them to do that. There are only two ways you can return cash to stockholders. One is to pay dividends, either regularly every period (quarter, semiannual or year) or as special dividends. The other is to buy back stock. From the company’s perspective, the aggregate effect is exactly the same, as cash leaves the company and goes to stockholders.”
8 Likes