We take on the problems that are supposed to be unsolvable.
Steelman is a small consultancy in Calgary. We are brought in when something important is stuck — a deal that will not close, a lender losing patience, a business whose own numbers are arguing against it — and the conventional answer is that nothing can be done.
Most “unsolvable” problems are badly stated.
When something cannot be solved, it is usually because the question is wrong, the evidence is unreadable, or the people who need to agree are looking at different versions of the truth. Those are not immovable facts. They are conditions, and conditions can be changed.
Our background is in building systems — automation, data pipelines, AI that does real work rather than demos. That matters here because the bottleneck in this kind of analysis has always been effort. Reading ten thousand pages, reconciling four conflicting statements, rebuilding a model from scratch when an assumption changes: those used to cost weeks, so nobody did them properly.
They now cost hours. That is the whole reason this service can exist at a price that makes sense — and the reason the work is more thorough than what it replaces, not less.
What we are actually good at
Forensic reading of bad documents
Scanned PDFs, workbooks with broken formulas, statements whose own subtotals disagree. We extract, reconcile and prove the numbers before we build anything on them — and we tell you what we found, including the parts that do not help you.
AI applied where it actually holds
We use it to read thousands of pages, to cross-check arithmetic, to catch the line item nobody noticed. We do not use it to generate conclusions. Every figure that reaches your document is traceable to a source.
Automation that removes the grind
The generators, extractors and reconciliation checks are rebuilt for each engagement and then kept. That is why a document that would take a consulting firm six weeks takes us days.
Arguments built to survive scrutiny
The counter-argument goes on the page, not in a footnote. Assumptions are adjustable by the reader. A case that only works when nobody touches it is not a case — it is a hope.
Rules we do not break
We build persuasive documents. That only works long-term if they are also true — the moment a lender catches one overstatement, everything else you have shown them is worthless.
- 01
The baseline has to be honest or nothing else counts
Our model reproduces your filed statements to the dollar before it does anything else. If a scenario looks better than reality, it is because of the decisions in it — never because we changed the basis underneath it.
- 02
Modelled is labelled modelled
Projections are badged as projections everywhere they appear. We have never presented a scenario as a restated historical result and we will not start.
- 03
We will talk you out of it
More than once we have looked at a situation and told the client their number was not supportable. That is the service working, not failing. You are paying for a case that holds, and the fastest way to lose one is to bring a weak one.
Tell us what is stuck.
The first conversation costs nothing and usually tells us both whether there is a case worth building.