Skip to content
DaylitedSimply Awesome

AI and robotics

Intelligence that has to be right, and machines that have to be safe

We are AI consultants and innovation consultants. We build the systems that answer for their output, and the cells that people can stand next to.

Start a conversationWhat it costs

4risk tiers, and the use decides which one
0of your data leaves the systems in the contract
1stop rule, and without it nothing goes to production

What an AI consultant is for

A model is easy to obtain and hard to be responsible for. Anyone can call an interface and get an answer back. What takes work is knowing whether the answer is right, what happens when it is not, and who is accountable when a customer acts on it.

That is the whole job. We take a question you have, find out whether a model is the right instrument for it at all, and then build the thing around the model that makes it usable: the evaluation set, the fallback, the logging, and the rule that stops it.

The answer is sometimes that you do not need a model at all. A lookup table, a threshold or a better form solves a surprising share of the work that arrives under that heading, and all three are cheaper to run and easier to defend in front of a customer.

An agent loop of four boxes with a stop rule drawn below it Four boxes named trigger, plan, act and check stand in a ring with arrows running clockwise between them. The check box is drawn heavier than the other three. To the right three violet boxes marked tools connect to the ring. From the check box a glowing line falls to a blinking point above a violet line, under the words a loop without a stop rule is not in production. TRIGGER PLAN ACT CHECK TOOLS A LOOP WITHOUT A STOP RULE IS NOT IN PRODUCTION TRIGGER . PLAN . ACT . CHECK

Agents

A loop without a stop rule is not in production

An agent is a loop: something triggers it, it plans, it acts through tools, and it checks the result. The first three are the easy part and the demonstrations all stop there.

The check is where the engineering is. What counts as done, what counts as wrong, how many times it may retry, and what it does when it has run out of retries. A loop that cannot decide to stop will eventually spend your money or your reputation on a task it misunderstood.

Two sets of bars, one from a demo set and one from a set the client owns Two labelled boxes stand side by side, demo set on the left and your set on the right, with an arrow between them. Under each stands a row of five bars. The left bars are taller and sit under a blue line marked looks done. The right bars are shorter and sit under a heavier line marked is done, with the shortest bar carrying a blinking point. DEMO SET YOUR SET LOOKS DONE IS DONE THE FAILURE YOU CAN NAME IS THE ONE YOU CAN FIX EVALUATION
System i drift
A corridor in a data hall. The model is the small part, and the chain around it is what has to be measured.

Evaluation

Your evaluation set, or the score means nothing

A score on a public benchmark says how a model does on somebody else's problem. It is worth reading and it is not evidence about yours.

So the first thing we build is usually not a model. It is a set of your own cases with the right answers written down, including the awkward ones, and a way to rerun it. Once that exists every later decision has a number behind it, and you can rerun the whole thing yourself after we have gone.

The failures get enumerated before launch rather than discovered after. A failure you can name is a failure you can put a fallback behind.

Three subjects

Where the work usually lands

Each of these has its own page, because the engineering in them is not the same.

An agent loop of four boxes with a stop rule drawn below it Four boxes named trigger, plan, act and check stand in a ring with arrows running clockwise between them. The check box is drawn heavier than the other three. To the right three violet boxes marked tools connect to the ring. From the check box a glowing line falls to a blinking point above a violet line, under the words a loop without a stop rule is not in production. TRIGGER PLAN ACT CHECK TOOLS A LOOP WITHOUT A STOP RULE IS NOT IN PRODUCTION TRIGGER . PLAN . ACT . CHECK

Agents and automation

Loops that do work: read, decide, act through tools, and check. The stop rule is the part that makes it a product.

The loop
A frame with three detections beside a bar chart cut by a confidence threshold On the left a frame holds three boxes, two drawn firmly and marked with high scores and one drawn faintly in violet marked with a low one. On the right four bars stand on an axis. A blue line crosses them marked threshold. The two bars below the line are violet, and one carries a blinking point, under the words what falls under it is still there. 0.97 0.88 0.41 WHAT IT SEES 1.0 0 THRESHOLD STILL THERE DETECT . SCORE . DECIDE . ADMIT THE MISS

Computer vision

Detection with a confidence score, a threshold that somebody chose, and an honest account of what falls under it.

What it sees
Two paths from sensor to action, one through the cloud and one on the device The upper row runs sensor, cloud, action with a measurement spanning it marked one hundred and eighty milliseconds and only when the link is up. The lower row runs sensor, on device, action, drawn heavier, with violet arrows and a measurement marked nine milliseconds, always. A glowing line joins the two rows at a blinking point. THE ROUND TRIP YOU DO NOT TAKE SENSOR CLOUD ACTION 180 ms . AND ONLY WHEN THE LINK IS UP SENSOR ON DEVICE ACTION 9 ms . ALWAYS THE MODEL SHRINKS UNTIL THE BUDGET FITS

AI at the edge

Inference in the device, where the latency budget is fixed and there is no round trip to fall back on.

The budget
Four stacked risk tiers with the second one drawn heaviest Four bars are stacked from the top down, marked unacceptable and banned, high risk and assessed, limited and disclose, and minimal and free. The second is drawn heavier than the rest and a glowing line runs from it to a blinking point. A measurement down the left is marked risk. Along the bottom stand the words the use decides the tier, not the model. UNACCEPTABLE . BANNED HIGH RISK . ASSESSED LIMITED . DISCLOSE MINIMAL . FREE RISK THE USE DECIDES THE TIER, NOT THE MODEL FOUR TIERS

The rules

The use decides the tier, not the model

European AI rules sort systems by what they are used for. The same model can sit in a minimal risk product and in a high risk one, and the obligations are completely different.

That is settled before launch, in writing, as part of the technical file. Doing it afterwards means discovering in a procurement questionnaire that the product needs a conformity assessment nobody planned for.

The declaration is the provider's own, which in almost every case is you and not us. We produce the evidence and the file it rests on, and we say plainly when a decision needs a lawyer rather than an engineer.

A robot cell seen from above with its safeguarded space and reach circle A violet rectangle marks the safeguarded space with a dashed rectangle inside it. In the middle a robot arm stands at the centre of a large dashed circle marked reach eleven hundred and twenty. Down the left a blue line of points is marked light curtain, and at the top right a circle with a blinking centre is marked stop. SAFEGUARDED SPACE REACH 1120 LIGHT CURTAIN STOP THE STANDARD DRAWS THE FLOOR PLAN

Robotics

The safety standard draws the floor plan

A robot cell is designed from the safeguarded space inward. Reach, speed and what happens when somebody opens a gate come first, and the layout follows from them.

That is a whole subject and it has its own page.

Or write here

Three fields, and a person reads it

Name, an address we can answer to, and what you are trying to build. You get a reference back straight away.

We use this to answer you and, where it fits, to prepare a scope. The lawful basis is steps taken at your own request before a contract. It is stored on our own server in the European Union, never passed to anyone, and deleted after twelve months if it does not become an engagement. The full notice is on the legal page.

An agent loop of four boxes with a stop rule drawn below it Four boxes named trigger, plan, act and check stand in a ring with arrows running clockwise between them. The check box is drawn heavier than the other three. To the right three violet boxes marked tools connect to the ring. From the check box a glowing line falls to a blinking point above a violet line, under the words a loop without a stop rule is not in production. TRIGGER PLAN ACT CHECK TOOLS A LOOP WITHOUT A STOP RULE IS NOT IN PRODUCTION TRIGGER . PLAN . ACT . CHECK

What would you point a model at?

Bring the question rather than the technology. We will tell you whether a model is the right instrument, including when it is not.

Start a conversationRobotics