You do not sell software. You sell hours — and most of them are spent waiting.
The deliverable assembled by hand for the fortieth time. The status report rebuilt every Friday. The regulator PDF re-attached across two messengers because no shared link for it exists. On a fixed-fee engagement none of that is billable and all of it is your margin. Below is what the work looks like today, what it looks like once Silow has mapped it, and what it gives back — recorded on the tools your people already use, with no keystroke logging and no scoring of anyone.
How the work looks today. How it looks after.
Seven processes, in the order we would take them. The left-hand column is not a caricature of a professional-services firm — it is what a recorded firm was observed doing. Rows tagged "Measured" carry a number that came out of real recorded work; rows that were not measured say so.
The work that goes into a deliverable is scattered across six windows: the file manager, two messengers, the mail client, the portal, the spreadsheet. A senior person spends the first half of the engagement finding, renaming and re-collecting the evidence for a document only they can actually write. In the recorded firm we reconstructed the real procedure it performs from ~88,000 scenes, and the assembly work is most of it.
The evidence for the deliverable is collected from what already exists across the tools: documents gathered, named to your convention, filed, gaps listed. The professional starts at the judgement — the part the client is paying for — instead of at the filing. Every draft is reviewed by a human before it reaches a client.
One regulator PDF was manually located in the file manager and re-attached across two messengers on five separate occasions over three weeks. No shared link for it existed. Every standard client request is retyped from scratch by someone who knows exactly what it will say, and has typed it forty times this month.
Standard requests are drafted from the language your team already uses, with the right attachment already attached. A human reviews and sends. The file has one retrievable place, so it is never hunted for again. This is the cheapest hour on the page to get back.
Every Friday, the same ritual: pull from the tracker, pull from the threads, remember what actually happened, rebuild the document, notice the thing that looks wrong, chase it. Fixed sources, fixed shape, fixed output — done by hand, by someone senior enough to know what the numbers mean, and billed to nobody.
The pull, the assembly and the format run themselves. What reaches the partner is the part that needs judgement: what moved, what slipped, what the client needs to be told carefully. This is the most repeated ritual in a firm that sells hours, and the clearest agent specification we produce.
A new client means the same packet assembled again from scratch — the intake questions, the document list, the first request, the folder structure, the invoice. The firm has done it forty times and it is written down nowhere, so the last person to do it is the process.
The onboarding packet is pre-assembled from the engagements that actually ran, in the shape the firm actually uses. The partner reviews it and sends it. Onboarding stops being a tax on winning new work, which is exactly the moment a firm can least afford one.
Work that needs a second human waits a median of 7 days. Work one person can finish alone waits 1.1 — a factor of 6.4. Two thirds of all waiting, 67.4%, sits across a handoff, and the client is one of the humans in that sentence. It is not time zones: the least-overlapping pair in the recorded firm still shares a quarter of the working day. The handoff is simply not defined anywhere.
Every handoff gets a shape: what is attached, what "done" means, who is waiting and since when — including when the thing you are waiting for is on the client’s desk. Nothing sits silently. This is the single largest block of recoverable elapsed time in the firm, and the cheapest one to change.
Three amounts sat in the trace — an invoice for 35,000, a figure of 35,700 EUR, a conversion of 122,000 EUR — and they do not reconcile. The firm was looking at them in turn, on different days, in different windows. The billing narrative is reconstructed on Friday afternoon from memory, which is the least reliable ledger in the business.
What happened, when, on whose desk and against which engagement becomes a by-product of the work rather than a task someone has to remember on a Friday. And when the trace cannot reconcile something, it says so: our report wrote only that the amounts were looked at in turn, and refused to invent the reconciliation. A system that will not fabricate is the only kind you can put near a client bill.
A firm’s asset is who knows what, and it lives in nobody’s system. In the recorded firm, after clustering the raw topics into a taxonomy, 26 working topics cleared the threshold and five had a single owner — one person is the only one who touches them. In a nine-person firm that is not a staffing note, it is the firm’s risk register.
Key-person concentration becomes visible at the level of the topic, not the person. The gate is deliberately harsh, and it worked against us: a topic holding 398 minutes of one person’s attention got no owner at all, because a single burst is not durable ownership. Silow does not score people. It shows you where the firm is one resignation wide.
The recordings behind this page: a 9-person firm doing licensing work, regulator documents, client onboarding and invoicing, run largely out of messengers — five weeks, ~325 hours recorded, 748 tasks reconstructed, ~88,000 scenes. And an 8-person cross-border operations team — 8 days, ~95 hours, 461 tasks. Both anonymised. Rows marked "Measured" come from that recorded work; rows we did not measure are marked as observed in the trace, or as a pattern rather than a measurement. Every figure was computed before a model was allowed to write a sentence about it, and all thirty supporting quotes in the reconstructed procedure were verified against the raw scenes they came from — not against a summary of them.
What an agent may touch, and what it may not.
In a firm whose product is professional judgement, this line is not an efficiency question — it is the question of what you are selling. So we draw it explicitly, before anything is built.
- Assembling the client deliverable’s evidence: documents gathered, named, filed, gaps listed.
- The recurring client status report — pulled, assembled and formatted from what already exists.
- The document chase: the file that gets re-found and re-attached, and the request that chases it.
- Drafting the standard client request your team has already sent forty times this month.
- The onboarding packet for a new engagement, pre-filled from the engagements that actually ran.
- The timesheet reconstruction: what happened, when, on whose desk, against which engagement.
- The advice itself, and the professional judgement the client is actually paying for.
- Anything carrying a signature or professional liability behind it.
- The client relationship, and every difficult conversation inside it.
- Scoping and pricing an engagement. Silow never decides what a piece of work is worth.
- Every draft an agent produces is reviewed before it reaches a client. Nothing auto-sends.
Everything on the left is the work around the advice. Everything on the right is the advice. Silow removes the assembly, the chase and the retyping, and leaves the judgement — and the named professional who is accountable for it — exactly where the client and the regulator expect to find them.
Margin, capacity, and the people you already have.
The rule for this section: not one invented percentage. Every figure below is either measured in a recording or supplied by you — your loaded or billable rate, not a benchmark we made up. A saving you cannot defend to a partner meeting is not a saving, it is a slide.
We do not quote you an industry percentage. Every candidate above is scored in hours against your recording and ranked by payback, so the first thing you build is the one that pays back fastest rather than the one that demos best. The anchors we can already point at: a median flow efficiency of 3.4% across 105 reconstructed tasks. A median seven-day wait every time work crosses a desk. A document re-found and re-attached five times in three weeks. 190 minutes over three weeks spent retyping threads into a system of record, by one person, on one ritual.
On a fixed-fee engagement the price is already agreed, so every non-billable hour comes straight out of the margin — and the queue is where those hours live. A task moving 3.4% of the time it exists is a fee you have already quoted, being consumed by waiting nobody has a line item for. On time-and-materials the same hours show up as utilisation: the assembly work is real effort your people cannot bill for, so removing it raises the billable share of the same headcount without anyone working longer. Both lines are hours × your rate — your rate, not one we invented. There is a third nobody counts: the AI licence you already pay for. In a recorded cross-border operations team — not a professional-services firm — the company had built its own assistant and recommended it; recorded usage across eight people was seven at zero percent and one at one.
No new system to adopt, no data to move, no process change while we look — Silow runs on the tools the firm already uses. The capacity comes back into the same team, doing the same work with the assembly removed: more engagements per partner, or the same engagements done better. Silow maps work, not workers. It produces no productivity ranking, no performance evaluation and no automated decision about anyone’s job — not as a policy written afterwards, but because the unit of analysis is the process. What you do with the released capacity is a decision for the people running the firm.
A task is moving 3.4% of the time it exists. On a fixed-fee engagement, the other 96% is not a scheduling inconvenience. It is your margin, sitting in a queue.
From a procedure nobody has seen to a layer you own.
If a deliverable can be assembled from a trace, the firm that owns its own trace wins — and the one that does not ends up reselling somebody else’s model at somebody else’s margin. The agents are the visible part. The record underneath them is the asset.
There is a document describing how the firm works and it is wrong. The real procedure — the one performed at 09:00 on a Tuesday, across a messenger and a file manager — has never been observed, which is why every automation attempt starts with a workshop and a guess.
In the firm we recorded we rebuilt the real procedure it performs, and all thirty supporting quotes were verified against the raw scenes they came from rather than against a summary. Ranked by impact, effort, risk and payback against your own hours.
The deliverable assembly, the status report, the document chase, the onboarding packet. Each arrives as a specification an engineer can build from: the observed steps, the scenes behind them, and the lines it must not cross. Not "this looks automatable" — the actual procedure, observed.
The agents are consumers of something more valuable: a private, structured record of how your firm actually delivers. That record is the difference between a firm that productises its own method and a firm that rents someone else’s. Onboarding, search, quality review, and whatever you build next, on data only you own.
Where AI actually works in professional services — and where it stalls.
Everyone selling AI to firms that bill by the hour arrives with a case study and a percentage. We are not going to quote you someone else’s, because we cannot verify it. Here are the patterns instead, including the failures.
The clearest win, and the least glamorous. Firms that industrialise the evidence-gathering in front of a document get more engagements out of the same partners. Firms that ask a model to produce the advice itself quietly stop after the first client review.
The differentiator is finding the last engagement that worked, not writing a new page from nothing. Where a firm has a real library, the payback is immediate. Where it does not, the model produces landfill at speed.
The highest-frequency, lowest-judgement work in the firm — and the one most teams believe they solved years ago with a template. It looks solved. It is a senior person’s recurring half-day, every week, billed to nobody.
Not a tool category, so no vendor sells it. Yet two thirds of the waiting in the recorded firm sat across a handoff between two people — and on a fixed fee, that queue is the margin. It is the largest number on this page and the cheapest to move.
Every firm knows it has one. Almost none can name the topics. Five topics in the recorded firm had a single owner — and that is a fact about the firm’s risk, not a score about a person.
Up to 66% of screen time in one team we recorded. Client work, approvals, promises and scope changes all live in threads, and none of it lives anywhere the firm can reuse, audit or sell back to itself.
They do not fail because the models are bad. They fail because nobody could say which of forty candidate processes to do first, or what "correct" looked like when it was done. That ranking is what Silow produces; the agents are what you build on top of it.
If the deliverable can be assembled from a trace, someone will assemble it. The firm that owns the trace of its own method productises it. The firm that does not becomes a reseller of a model it does not control, at a margin it does not set.
The part other vendors leave out.
Two of the hypotheses we most wanted to be true were killed by our own data. In a business built on professional honesty, the vendor who tells you what their product cannot do is the one worth the second meeting.
The recording sees activity, never outcome. Whether the engagement was profitable, whether the client renewed — that is not in the data. Join one column of your own outcomes to our trace and it becomes provable. Until then, any ROI figure quoted at you is a guess, including one quoted by us.
The pitch was "N people doing the same thing, here is the saving." The real answer across nine people over five weeks was 44 recoverable minutes — 0.22% of working time, with a median duplicate of fifteen seconds. There is no product there, so we do not sell one.
Three quarters of what our naive metric called "meeting minutes" had no room attached at all. That is a capture problem, not a metric problem. We cut the meeting-ROI report rather than ship a chart that would flatter us.
The trace surfaced 24 live client threads that had never been entered into the system of record — while the same team was hand-building a browser extension to scrape their own leads in. But roughly four of those rows were genuinely hard. We would rather tell you that now than after you have bought it.
Weeks, not quarters.
Silow runs on the firm’s existing machines and tools. No migration, no new system, no process change while we look — the point is to see the work as it actually is. You choose the deployment mode, including fully air-gapped.
Scenes become tasks; tasks become the real procedure the firm performs. Numbers are computed before a model writes a sentence about them, and every quote is checked back against the raw scene it came from.
Every candidate scored by impact, effort, risk and payback against your own hours and your own rate — so the first thing you build is the one that returns margin fastest.
A specification, not a suggestion: the observed procedure, the evidence, the scenes behind it, and the boundaries it must not cross. The advice, the signature and the client relationship stay where they were.
What you do with the capacity is yours.
Silow maps work, not workers. No productivity ranking, no performance evaluation, no automated decision about anyone’s job — not as a policy written afterwards, but because the unit of analysis is the process. We are aware the buyer in this industry is often thinking about headcount; the product genuinely does not do that, and we would rather say so than let you find out. Two weeks of recording, and the first ranked roadmap lands with the hours attached.