On industrial projects, delay is tracked in weeks, never priced. Everyone sees the slippage; no one puts a figure on what it destroys, in margin, in deferred savings, in risk drawing closer. SEI turns that time into money: before you commit the funds, it states what a week of delay would cost; once the project is running, it sees the drift coming and prices it as it forms.
Demonstration · illustrative data from an anonymised portfolio · no client data
Five steps, from the file you already produce to the measured effect of the decision. Each one is verifiable and traces back to its source.
The exported schedule, the financial extract, the minutes, the investment case. Nothing extra to key in, no connection into your systems.
A milestone slipping, contingency drawn down faster than progress, spend lagging the profile of its nature. The signal arrives before the finish date moves.
The chain of milestones affected, the range in your own currency, the recalculated finish date, and what remains to be checked.
Continue, defer, resize, or fund a recovery. SEI prices all four options and names the authority that decides: project manager, sponsor or committee.
The decision is logged with its estimated cost. At close-out, the gap between estimate and actual feeds back into the model.
Not the offers, but what runs underneath them. Four building blocks turn your projects into priced decisions, and make the whole thing hard to copy.
Turns every week of delay into money, in your own currency: margin never produced, savings deferred, risk carried. Calibrated by project nature, never a hardcoded figure, always a range.
Calibrated on real, closed projects. It recognises the profile of those that drifted (resources, ownership, scope, communication) and scores a project underway before it derails.
Reads what your systems already produce (ERP exports, schedules, minutes) and extracts the facts that matter. It replaces no tool; it plugs the economic read-out onto what exists.
SEI doesn't train on the internet: it trains on your industrial history. The mechanism is simple and checkable: at the close-out of every project, the gap between estimate and actual is fed back into the model. Your data stays in your instance and never trains another company's model, so two companies never get the same one.
Industrial companies committing to long investments in regulated environments, where commissioning cannot be redone and where the promised value depends on a date.
You will recognise yourself if your committee weighs cases appraised under different assumptions. If your projects are rebaselined without anyone pricing the shift. If your estimating benchmarks live in three people's heads.
Trixelia builds SEI. One calculation, applied at three moments in the life of an industrial investment. Before committing, to state what the project is worth, what a week of delay would destroy, and what risk it carries. During execution, to price every slippage as meetings go by. At close-out, to settle the gap between what was promised and what was delivered.
These are not three products. It is what makes the promise hold: a cost of delay invented mid-project would be worth nothing. It has to be born in the investment case, locked at commitment, and settled at close-out.
One founder, three locations: London, Switzerland and France.
The reference portfolio the model is built on holds dozens of closed industrial projects, analysed case by case down to the root cause. It is anonymised: no company, site or individual is identifiable. That is the condition for yours to stay that way too.
A project delivered ninety-one weeks late. Reporting green all the way to commissioning. And at close-out, nobody around the table able to say what that delay had cost.
Going back through the closed cases one by one, the second discovery was more troubling than the first. Those projects had not only been badly executed: they had been badly appraised. Approved under conventions that changed from one case to the next, on estimates the site's own history already contradicted. A project approved below the benchmark for its nature ends up drifting all the way to that benchmark. It did not slip; it had been underestimated.
The same factors, read at the same milestones, produced the same drift. What looked like a run of bad surprises was a pattern, and a pattern can be modelled.
Trixelia came from there. Not from a hunch about artificial intelligence, from a finding in a set of files.
Founder and CEO
Annabelle Frisa
Over fifteen years in regulated industries, running capital projects where a mistake cannot be undone in production.
I spent fifteen years running industrial projects in environments where every decision is documented, justified and controlled. Everything there is measured: budget variances, non-conformities, qualification lead times. One thing is never priced, and it is the one that costs the most. Time.
A budget overrun has an owner. It has its line, its variance, its item on the agenda. A delay has none. You rebaseline, and it disappears. I have watched projects stay green all the way to commissioning and destroy millions in value without a single indicator lighting up.
And when I went back through the files, I found worse than delays. I found projects approved under assumptions that changed from one case to the next: a contingency rate running from zero to twelve per cent, a residual value assumed anywhere between zero and eighty. A committee that thinks it is choosing between two projects is often choosing between two sets of assumptions.
I built SEI because that information already exists. It is in the investment case, it is in the schedules, it is in the meeting minutes. Nobody had connected them.
What I want is that no industrial investment decision is ever made again without its cost having been priced.
Industrial investment decisions have never had a real intelligence behind them. We're building it: a model that turns projects into priced, defensible decisions.
We're looking for people who want to work on hard problems that matter, and shape how an entire industry decides. If that's the kind of work you're after, let's talk.
Build with us →