Operational Discovery Report
Westmore Partners — 30-person boutique strategy consulting firm
Prepared from 4 interviews · Conducted Four interviews across one engagement team, one partner, and one operations lead · Total interview time: 1 hours 21 minutes
What we found, in one paragraph.
The firm feels slow because a meaningful slice of every week is spent reconciling the gap between the systems Westmore bought and the way Westmore actually works. The deck template in SharePoint does not match what the partners want, so Maya rebuilds decks every Friday afternoon from a Notion doc only she maintains. The time-tracking system cannot export in the format the partner billing meeting needs, so Tom runs an Excel macro he built as a senior analyst that now silently powers three meetings. Bill.com, QuickBooks, and Tom's sheet never agree, so David spends Sunday night reconciling them and chasing mid-week billing-code changes that should have surfaced on Wednesday. Salesforce sits unused because it was built for pipeline-stage selling and Westmore sells through partner relationships, so Sarah triages prospects in her own Notion CRM for ten-plus hours a week. None of this is anyone behaving badly. It is four sensible people each absorbing the cost of a system that does not fit, and the firm has organised itself around their workarounds so completely that the workarounds are now the operating model.
The three things you should know first.
The firm's real operating system is three personal artifacts, and none of them are documented.
When we asked what would break if each person left, the answers all pointed at the same kind of thing. Maya has a Notion doc called 'What everyone wants' that records every partner's deck preferences — Sarah's 24pt Garamond headers, Tom's allergy to two-column layouts, the ex-McKinsey partner's pyramid-bullet exec summaries. Nobody else uses it. Sarah confirmed it independently in her interview: 'When a new associate joins my team I send them to Maya, not to HR.' Tom built an Excel macro two years ago as a senior analyst that the partner billing meeting, David's Sunday reconciliation, and the part-time CFO's monthly close all now depend on. As Tom put it: 'If my Excel breaks, three meetings break.' Sarah keeps her prospect pipeline in a Notion CRM that, in her own estimate, is sitting on roughly $2M in fee revenue over the next 18 months from informal commitments that exist in no shared system.
What is striking is that none of this is shadow IT in the usual sense. These artifacts are not hidden — they are central. The partner billing meeting on Wednesday is built directly on Tom's macro output. New associates are routed to Maya's FAQ. Sarah's Notion is the prospect tracker the firm runs on. The artifacts have become institutional infrastructure without ever becoming institutional. A 30-person firm has, in effect, outsourced three of its most important operational systems to three individual contributors' personal files.
The risk is not theoretical. Maya said the Friday deck rhythm would 'break for at least a month while someone learned which partner wants what.' Tom said David would 'have to learn macros, which — bless him — is not where his interests are.' Sarah said her informal commitments would 'lose the thread.' None of them was being dramatic. They were describing the operational reality.
*Three concentrated single-points-of-failure, each carrying somewhere between two and twelve months of recovery time if the person left. Difficulty to fix: medium, but mostly organisational rather than technical.*
Every official system has been quietly routed around, and the routes are the actual workflow.
Westmore bought a SharePoint deck template, Salesforce, and a time-tracking system. In practice, the firm uses none of them as intended. The SharePoint template exists but, in Maya's words, 'half the partners don't use it because it doesn't have the slide layouts they like.' So she rebuilds decks in each partner's 'actual' template, 'which only exists in their head and on their last project's files.' Salesforce was bought three years ago and has not stuck. Sarah was unusually direct: 'Salesforce wants pipeline stages and probability percentages. We don't have a pipeline. We have a Rolodex.' The time-tracking system cannot export in the format the partner billing meeting needs, so Tom's spreadsheet became the export. The firm has tried to replace the time-tracking system three times. Each attempt has stalled — and Tom put his finger on exactly why: 'Nobody can articulate the export format the partners need, because the partners can only describe it by pointing at my spreadsheet.'
This is the same pattern as Finding 01, viewed from the other side. The personal artifacts exist because the official systems do not fit. The official systems do not fit because the firm's actual operating logic — relationship-driven sales, partner-specific deck preferences, partner-defined billing breakouts — was never the logic the off-the-shelf software was designed for. Buying more software will not solve this. The next tool will be routed around for the same reasons.
What this means in practice is that Westmore needs to either explicitly retire the official systems that nobody uses (and codify the personal artifacts as the canonical ones) or invest in adapting the official systems to how the firm actually works. Today the firm is paying the cost of both — software licences for tools nobody uses, plus the human cost of the workarounds. Picking a side is upstream of almost everything else in this report.
*An estimated 15–20 weekly hours of duplicate work across the firm trace directly back to this pattern. Difficulty: low to make the decision, medium to execute on it.*
The week ends on Sunday because mid-week information does not propagate.
David's Sunday-night reconciliation was the most vivid version of this. He spends four to six hours every Sunday reconciling Bill.com, QuickBooks, and Tom's master sheet. The three never agree. The reason they do not agree is almost always a billing-code change that happened on a Wednesday — most often Sarah re-categorising an engagement from 'advisory' to 'implementation' because the scope shifted — that nobody told David about. David hears about it on Sunday at 9pm when his numbers do not match. He emails Sarah, gets the explanation Monday, fixes it, and the week starts again. As he put it: 'Mostly because of when it is, not what it is. If the billing-code change conversations happened mid-week instead of being discovered Sunday at 9pm, my Sunday could go back to being a Sunday.'
The same shape shows up elsewhere. Tom spends Tuesday morning chasing associates who did not file timesheets by Monday morning, because his macro calculated wrong recovery rates on Monday and the partners noticed. Maya gets last-minute polish work pushed onto her Friday — sometimes Saturday — because deck specs arrived late in the week. In each case the issue is not the volume of work. It is that the work is being discovered at the moment of failure rather than at the moment of change.
Westmore does not need more meetings to fix this. It needs a small amount of mid-week visibility — a Wednesday signal when a billing code changes, a Thursday signal when timesheets are missing, a daily signal on which decks are due Monday and whose templates they need to match. That kind of ambient awareness is exactly what a well-scoped agent does well, and several of the agents proposed below do exactly that.
*Roughly 8–10 hours per week that are currently happening on evenings and weekends could move back into the working week with better mid-week signal flow. Difficulty: low technically, medium culturally.*
The 8 high-value findings.
These are presented more briefly. Each was mentioned by multiple interviewees or surfaced as a pattern across teams.
**Maya's Friday deck polish is translation work, not analysis work.**
Maya spends four to five hours every Friday — sometimes into Saturday — rebuilding decks in each partner's preferred-but-undocumented template. Her own framing: 'It's not analytical work — it's translation work.' The translation layer is a senior analyst's Friday afternoon, every Friday, all year.
*~200–250 hours per year of senior-analyst time on a task that is not the work the firm sells.*
**Tom's timesheet macro is the de facto specification for the time-tracking system.**
Three separate attempts to replace the time-tracking system have stalled because the export format the partners want only exists implicitly in Tom's Excel macros. Until someone extracts that spec into a written document, the fourth attempt will stall too.
*Unblocking this unlocks the path to retiring ~6 hours of weekly manual reconciliation across Tom and David.*
**David's Sunday reconciliation is driven by Wednesday silence.**
Bill.com, QuickBooks, and Tom's sheet never agree because billing-code changes happen mid-week and surface on Sunday night. David has the patience and the chasing emails — but the conversations are happening at the wrong end of the week.
*4–6 weekend hours per week, recoverable if mid-week change signals exist.*
**Salesforce is being paid for and not used; the real CRM is Sarah's Notion.**
Sarah was unambiguous: Salesforce was built for a pipeline-stage model and Westmore sells via fifteen-year relationships. Every partner maintains their own version of a personal CRM. Roughly $2M of informal prospect commitments live in Sarah's Notion alone, with no shared visibility.
*~10+ hours/week of Sarah's time on triage, plus single-partner key-person risk on ~$2M of forward revenue.*
**Late timesheets cascade into Tuesday-morning fire-fighting.**
Tom's macros run on Monday morning's CSV export. If associates haven't filed by then, recovery rates calculate wrong and partners react. Tom spends Tuesday morning chasing instead of doing engagement work. A simple Friday-afternoon nudge would absorb most of this.
*~3 hours/week of EM time, plus partner-confidence cost in the Wednesday billing meeting.*
**Onboarding has quietly become Maya's unpaid second job.**
Every new analyst class asks the same questions in the same order. Maya built an FAQ that lives next to her partner-preferences Notion. Nobody points new hires at either. Sarah confirmed she sends her new associates to Maya rather than HR.
*Single-digit hours per onboarding event, but career-distorting concentration of institutional knowledge in one senior analyst.*
**Cross-partner prospect activity is invisible to anyone but the individual partner.**
Each partner maintains a private prospect tracker (Notion, email, or in their head). Partners share notes by forwarding emails. If two partners are talking to the same prospect, nobody outside the partners would necessarily know.
*Unknown but non-zero risk of duplicate outreach, lost referrals, and prospects that go cold during partner absences.*
**The brand template in SharePoint is the version nobody uses.**
The official template exists, but the working templates are partner-specific and unwritten — Sarah's 24pt Garamond, Tom's no-two-columns, the ex-McKinsey partner's bullet pyramids. Maya's Notion captures them; the firm's template library does not.
*A canonical, partner-aware template library is the prerequisite for any deck-related automation.*
What we'd build first.
Three pieces of upstream process work that should happen before — or alongside — the agent build. Each of them is a decision the leadership team needs to make, not a tool to buy. The agents in the next section become substantially easier and substantially more valuable once these are in place.
**Pick a side on Salesforce.** Either commit to making it work for relationship-led selling — which means custom objects, partner-specific views, and accepting that 'pipeline stage' is the wrong abstraction — or formally retire it and codify the partners' Notion CRMs as the canonical prospect system.
Right now the firm is paying for Salesforce and getting nothing from it while every partner runs a private system. Until this decision is made, any prospect-side agent has nowhere stable to live. This is the upstream of Finding 04 and of Agent 3 below.
**Reverse-engineer Maya's 'What everyone wants' Notion into a real, version-controlled partner-preference library.** Make it official. Move it out of Maya's head and her personal Notion and into a place the firm owns, with named maintainers and a process for updating it when a partner's preferences change.
This is the prerequisite for the Friday deck-polish agent. It is also the prerequisite for Maya being able to take a vacation. Doing this without an agent already produces value; doing it before an agent is built makes the agent much more reliable.
**Write down the export specification that currently lives only inside Tom's macros.** Sit with Tom for a few hours, document the exact billing-code grouping, phase breakouts, and recovery-rate calculations the partner billing meeting depends on, and produce a written spec.
Tom himself diagnosed this: every time-tracking replacement has stalled because the partners can only describe what they need by pointing at his spreadsheet. The spec on paper unblocks the time-tracking RFP, the timesheet reconciliation agent below, and the eventual retirement of three weekly manual processes.
None of these three is glamorous. They are also all explicitly things humans should decide, not things an agent should decide. Get them right and the agents below become straightforward implementations rather than political projects.
The agent blueprint.
For each of the high-value patterns above, a specific AI agent we'd build, in the order we'd build them. The order matters: the first agent is the trust-builder, the second depends on the spec work in priority 3, and the third is the highest-stakes and goes last on purpose.
Friday deck-polish agent
Timesheet reconciliation and mid-week nudge agent
Partner prospect Rolodex agent
A note on what we didn't find.
A few things worth saying, because reports like this tend to oversell.
We didn't find a villain. Everyone we spoke to is doing reasonable, conscientious work. Tom did not build his macro to entrench himself; he built it as a one-time favour and never got handed an off-ramp. Maya did not build her Notion preferences doc to become indispensable; she built it because the partners' preferences kept changing and she needed somewhere to write them down. Sarah is not snubbing Salesforce out of stubbornness; she is using the system that actually fits her sales motion. The dysfunction here is structural, not personal.
We didn't find evidence that headcount is the problem. The hours we identified are not 'we don't have enough people' hours — they are 'the same people are doing the same translation/reconciliation/triage every week' hours. Cutting headcount would make the workarounds collapse; adding headcount without changing the workflows would just produce more workarounds. The right move is to remove the work, not to add or subtract the people doing it.
We didn't talk to enough of the firm to claim this is exhaustive. Four interviews — one senior analyst, one engagement manager, one partner, one operations lead — is a representative slice, not a census. We did not interview the other partners, the part-time CFO, the rest of the associate class, or anyone client-facing about delivery quality. Several findings (especially the cross-partner CRM observation) are stronger in inference than in data and would tighten substantially with a second round of conversations.
We didn't find evidence that any of this is hurting client work directly — yet. The decks ship. The bills go out. The prospects mostly get responses. What the transcripts described is the cost of that being true: people working through Friday afternoons, Sunday nights, and Tuesday mornings to keep the visible work looking smooth. That cost is real, but it is also recoverable, which is the point of the agents above.