I build systems that do the expensive part of expert work.

Two strands so far. Investment diligence: reading a company's filings and its data room together, and returning something a partner can act on in days rather than weeks, with every claim traceable to where it came from. And automation cost: what a metered workflow stack actually costs a business, and what the same work costs once it isn't metered.

The fastest way to judge that is to read one. I pointed the engine at an AIM-listed company and froze every source the day before its takeover was announced, so the analysis can be checked against what actually happened.

The second strand has a worked example of its own. Rather than wait for a client's data, I ran the document I produce end to end against a constructed subject and published every price it depends on. The finding that surprised me: of seventy-two live workflows, the sixteen doing client reporting were one per cent of the bill, and four lead-routing families were forty-six per cent of it. That is invisible from inside a billing screen, and it means a migration priced per workflow is priced on the wrong unit.

What I build

I'm co-founder and President of Moneta Intelligence, where I built our diligence engine end to end. It takes a company name or a folder of confidential documents, reads the open internet and the private files together, produces five expert assessments and an independent valuation, fact-checks its own output, and returns a sourced report with a printable executive summary.

The same approach has become other things: a qualitative screener that reduces a large universe of companies to a shortlist on non-financial criteria before any quantitative filter runs; an earnings-momentum system built to a portfolio manager's own universe; a macro pipeline that reads what central banks actually publish and turns it into something a trading system can act on. I've also talked through how all of it works, at length, with someone who asked good questions.

The second strand is smaller and newer. Agencies and other businesses that run their operations on metered automation platforms pay per task, so the bill grows with the number of clients they serve rather than the work they add. I take the stack apart, cost every workflow, and say which are worth moving and which are not. The same standards apply: every price sourced and dated, and the arithmetic shown rather than asserted.

How I work

• ProvenanceEvery claim is traceable to its source, with an audit log behind it. If a number can't be sourced, it doesn't ship.
• ConfidenceFindings are labelled high, medium or low, and the report states plainly what the evidence cannot establish.

What this doesn't do

On public data alone, the engine deliberately stops short of a buy or pass verdict. Without confidential financials, issuing one would be dishonest, so it presents the intelligence and leaves the call to the person making it. It is a strong first pass, not a replacement for an audit or for a partner's judgement.

There is a second limit worth stating. A proof of cash, tying reported revenue to money that actually arrived, needs the bank statements. On a listed company you get an audit opinion but no ledger; on a private deal you get the ledger but no audit. Different assurance, not more of it.

On the automation side the honest answer is sometimes that there is nothing worth doing. The bill can be too small to repay the work, or the largest saving available can be a change of billing terms rather than a change of platform. The teardown says so when that is what the numbers show, and I would rather tell someone that than take the fee.