AI agency
You were looking for an AI agency. We are not one, and for most assignments we think that is an advantage. Here are the arguments, including the ones against us.
Short answer
An AI agency is an organisation that sells AI services with salespeople, project managers and consultants in several layers. Cekta is instead two engineers who do the whole job themselves: mapping, building, training and operations. You pay no overhead and you talk to the same people from the first call to daily production.
An AI agency sells AI services as an organisation: strategy, development, integrations and sometimes maintenance. The work is split across roles, usually a salesperson or account manager, a project manager and a number of consultants with different specialities.
The model has real advantages. An agency can staff several tracks in parallel, has someone to cover when a person falls ill and can scale up when a project grows. For large programmes with many simultaneous parts it is often the right way to work, and then you should choose an agency.
It also has built-in costs, and they are paid by the client. Few people are open about them, so we will be.
Four things that come with the organisation, however good the people in it are.
01
Salespeople, account managers, project managers, offices and marketing sit in the hourly rate. It is not a trick, it is what running an organisation costs. But you pay for it, and you do not get it back in the form of code or mapped workflows.
02
The most common pattern in consulting: the most experienced person sits in the sales meeting, and the following week a different team does the work. Ask who will actually write the code, and ask for names. There is a good answer, but it has to be said out loud.
03
Every time an assignment changes hands, from sales to project management to development to maintenance, something that was said in the room disappears. The more layers, the more of the business understanding that has to be retold instead of simply staying with the person building.
04
An organisation with employees needs to keep them booked. That makes nobody evil, but it does make it harder to recommend stopping. A project that should have ended after the pilot often runs another quarter, and it is rarely a decision anyone makes consciously.
Cekta is Caesar Katende and Ted Wachtmeister. Caesar works on agent architecture and orchestration, Ted on backend, data and ML as well as operations. Both studied at Lund University, and both are in every assignment from the first call.
The same people all the way: the person who draws the strategy writes the code and runs the training. Nothing gets lost in a handover.
There is no sales organisation to get through and no project manager who has to check internally before you get an answer. You get a fixed price before we start, one bounded workflow in production in 2-4 weeks and the code in your own repositories.
Four concrete differences, not an attitude.
01
We bound the assignment until we can set a price we stand behind. That puts the risk of the project growing out of control with us and not with you, and it forces us to understand the task before we start rather than along the way.
02
One bounded workflow built all the way to production in your existing systems. Not a demo, but something that runs for real and that you can base the next decision on. The effect can be measured before you have committed to a year.
03
Code, documentation and configuration are left in your own repositories. You can change supplier or take over operations yourselves without negotiating about what you have already paid for. Depending on us should be a choice you make every month.
04
The person who answers the first call is the one who builds it and the one who runs the training afterwards. It makes the answers faster and considerably less diplomatic, and you get told that something is difficult while it can still be changed.
Three legs, the same two people in all three.
01
A mapping of your workflows and a prioritised plan with indicative prices and a timeline per initiative. If you move on to a pilot within three months, the price of the mapping is credited. The indicative prices are open on /en/ai-strategi.
02
The build: one bounded workflow in daily use in your existing systems, with a human left to review before anything becomes binding. Operations and maintenance afterwards if you want them. Indicative prices on /en/ai-implementering.
03
Talks and workshops in the tools you already pay for, with exercises taken from your own work and a block on data hygiene that is always included. Indicative prices on /en/ai-utbildning.
We publish indicative prices instead of asking you to get in touch for a quote. A simple automation is SEK 30,000-60,000, a pilot in production SEK 60,000-150,000 and an automation mapping SEK 25,000-45,000. The full tables are on /en/ai-implementering and /en/ai-strategi.
The price is fixed and set before we start. Comparing it with an agency's hourly rate is difficult, and we do not think the comparison belongs there. Ask instead what you get into production for a given amount, and who does the work for the money.
We do not take assignments that need ten people in six parallel tracks. We say so in the first call instead of staffing up and hoping it works out.
We do not do advertising, campaigns or content production with AI. There are agencies that are genuinely good at that, and we are not one of them.
We do not hire ourselves out as full-time consultants inside an internal team. We deliver bounded assignments with a result, and then operations if you want them.
Common questions
When the work has to go wide at the same time: several parallel tracks, many stakeholders and an organisation that needs staffing up for a year or more. Then the agency structure is an advantage and not a cost. If we think that is your situation, we say so in the first call.
It depends on the project, and we answer honestly in the first call. One bounded workflow in production we handle without trouble, as we do several workflows one after another. A programme with ten parallel parts we do not do, and then the answer is a no, ideally with a tip about who you should talk to instead.
You do not, and the same goes for an agency. That is why we build so that you can manage without us: the code is yours and sits in your repositories, everything is documented and operations can be handed over to your IT function or to someone else. Lock-in should not be the thing that keeps you.
There are two of us and we work in each other's code, so a shorter absence affects the pace but not the delivery. For a longer absence we tell you straight away and agree a new timeline instead of letting it drift. It is also one reason we keep assignments bounded.