blog

Special services for your department

The time has comethat we need to address the repeating question from our partners, customers, friends and peers: what is the master plan of @KnowDroids.ai? What is our technological goal (in addition to our purpose - which we elaborated on our Linkedi)? 
The world is full of buzz of the kind of “is this or that thing already sentient”, “AGI is around the corner”, “what happens to us when AI gets smarter than even the 10th percentile of human experts”. In the meantime, the majority of humankind is still happily using pen and paper. In our own business circles, we see customers, some even the largest companies in the world,  handling tons of Excel sheets, filling them manually and then copy-pasting from somewhere to somewhere else, appending them to emails and sending them to someone who for sure needs to read them. All this is making me somewhat sure that AI will not prevail anytime soon; we ought to make it just reasonably useful!
And that is our technological goal: making our AI agents just reasonably useful. With this reasonable usefulness we serve a number of large companies, helping them automating enterprise-wide administrative tasks in the fields of compliance, market intelligence, reporting or customer experience and care. Besides this, we are also preparing a suite of light-weight and simple to use agents for so called professions - finance professionals, compliance experts, internal audit specialists or lawyers. Technologically, we of course face a number of familiar challenges: how to deal with the multi-agentic topology? What short-term and long-term memory options would be the optimal ones? How to make our customers’ legacy systems agents-ready? These are many, andalways will be, but when it comes to the core of our master plan, it distills into just a limited number of guiding principles which are all-premeating our agentic road map. Essentially, the following three:

We build AI agents, that are

  • Fit for purpose,

  • Efficient,

  • Fristonian.

Sounds simple? Well...

Agents that are fit for purpose

Our customers and users are having certain expectations and utility in mind, and we construct agents that deliver that utility. Not a proximal utility, nor something surprisingly good, but adjacent to or different from what was the original expectation - our agents ought to deliver exactly what they were composed for. In practice, “meeting the expectation” is a goal of paramount size and height. 

Agents that are efficient

Imagine you have an agent delivering your expectation already, doing your task. What is the next stage then? You certainly want the quickest possible agent doing that. Or the least expensive. Or the most secure. Or one which deals with your systems, not with some ideal systems you don’t currently possess. Or… You name it. Efficiency is a quality which is not found in LLM and agentic benchmarks, but which - believe me, I have been there - a serious business customer requires right alongside the essential utility of the solution.

Agents that are Fristonian

For what I mean by this, you may want to get familiar with the work of Karl Friston - a soft introduction I wrote again on our Linkedin. If I simplify a lot, I will call Fristonian such an agent, which learns over time and explores and acts in his (her?) environment in the way that optimizes his action in a certain way over time. Humans may be considered Fristonian agents in a certain sense, too. Collaborative systems - such as organizations consisting of humans and agents - can as a whole be at the end Fristonian. The “Fristonianity” goal spans scales from small to large, from simple to very complex - and some of our agents we will, over time, make, initially modestly, Fristonian, meaning less static and to an extent self-improving.
On the actual underlying technology we are developing at @knowdroids.ai, beyond the master plan principles, we will write morenext time.
Droidsfully Yours,

Vladislav Severa

CEO, KnowDroids.ai

KnowDroids.ai

Jindřišská 941/24,
Praha 1, 11000,
Czechia