A partner asks the memory four questions from her phone before a call, and no software drafts anything.
Whitlow Ferris Partners is a fictional management consulting firm that works for private-equity portfolio companies. It runs Microsoft 365, BigTime, HubSpot, and QuickBooks Online, and its method and positions have been organized and approved. Here the managing partner, twenty minutes before a call with a repeat sponsor, asks the memory four questions from her phone. The answers come back with their sources, nothing is drafted or generated, a fifth question gets an honest refusal, and she decides what to take into the call. Each step is the screen a real person would be looking at.
The value in this workflow is how long a partner spends before a sponsor call finding what the firm already knows, and how often the call is taken on memory alone. We would measure both on your own sponsor work. The diagnosis measured proposal development and not this workflow, so we do not put a number on it.
WFP Knowledge · in Teams
Ellen asked the same questions before every sponsor call, and they went to three people by Teams message.
Jen Halvorsen answered for the sprint, Marcus Delgado for what Dan reads first and for the carve-out position, and Carla Stimson for realization. The answers arrived when those three came out of their own meetings, which was sometimes after the call had started, and what Ellen remembered from the last Larkspur call filled in the rest.
The form sits in Teams on her phone. It is a question form and a search over the firm's memory, and it does not draft anything or summarize the call for her.
- Who is here
- Ellen Marchetti, managing partner. Between meetings, on her phone.
- What is running
- A question form and a search over the memory, with no drafting, no generation, and no decision made by software.
- Where it reads from
- The firm's memory, in the WFP Knowledge library on SharePoint, inside the firm's own Microsoft 365 tenant.
9:40Ellen has typed four questions into a form that reads from the same memory the firm's other software reads from, and nothing has been drafted.
Answers cite where they came from
ENG-2611 · engagement record
Every answer names the entry it came from, so Ellen can tell an approved position from a kept record.
The first answer is a kept record: three findings from the August readout in three lines, and the sprint status from the engagement record. Both are marked as records, so Ellen knows they describe what happened and are not a firm position.
The second answer changed yesterday. The debrief of the Beltline diagnostic proposed a risk page for Larkspur Ridge readouts, the addition to SP-2 was reviewed and approved on 23 September, and the answer Ellen reads at 9:44 carries that date. Before the memory, the preference lived in Marcus's inbox and the change would have lived in the Lessons channel.
- Elapsed
- Four minutes, between meetings.
- Before this
- Ellen would have messaged Jen Halvorsen and Marcus Delgado, and the answers would have arrived when their own meetings ended.
- Who decided
- Ellen. The memory answered two questions. It did not summarize anything for her and it did not tell her what to say.
9:44Two questions get two answers, each one naming the entry behind it, and the second carries the date it last changed.
Answers cite where they came from
Source: the Monday brief of 21 September.
The third answer stops at what BigTime can support.
Realization under UT-1 is fees billed over the standard-rate value of hours charged, so it exists only once an engagement has billed. The two closed Larkspur Ridge engagements this year were led by Ellen and Marcus and realized in the low-to-mid nineties. The third, at Tolland Wire Products, is Rob Askew's, at week two and running eleven percent over on hours, and the answer says its realization is not yet knowable, with no projection offered.
A tool that projected a figure here would look more helpful and would be wrong often enough that Ellen could not rely on it. Because it does not project, she can repeat what it says about the two closed engagements to a sponsor without checking first.
- The definition
- UT-1, owner Carla Stimson. The same definition the Monday brief and the diagnosis use, so the figures agree across the firm.
- The source
- BigTime, realization by engagement, as read by the Monday brief of 21 September. The overrun at Tolland was flagged there with Rob as its owner.
- Who decided
- Ellen. Whether the Tolland overrun comes up on the call is her judgment; the memory gave her the status and stopped.
9:48The third answer gives what BigTime can support and says the rest is not yet knowable. The refusal to project is written into how the memory answers.
Answers cite where they came from
· ACC-LARK, account summary: five engagements in three years, contact Dan Kowalczyk.
· The Monday brief of 21 September: one Larkspur Ridge invoice, Midland, fifty-one days old at $64,000.
The fourth answer was recorded six days earlier, when a consultant asked the same question.
On 18 September Kyle Brandt asked the playbook desk the firm's view on carve-out pricing. Nothing was documented, the question went to Marcus Delgado once, and his answer came back as FR-3. Ellen reads the same entry from her phone, with his name and the date on it. The playbook desk sample shows the question being asked.
The fifth question has no entry behind it, and the answer says so, with the two related records it does have marked as not being a position. Ellen decides to ask Dan on the call. Whatever he says goes to Marcus to record, and the next person who asks gets an answer with a date on it.
| Beltline | Three findings from the diagnostic; the sprint is at week four, and the steering committee met yesterday with the readout opening on the value bridge. CASE-2611, ENG-2611. |
| How Dan reads a readout | The value bridge first, the deal team's three KPIs on page two, and a risk page after the bridge since 23 September. SP-2. |
| Realization on Larkspur work | Low-to-mid nineties on the two closed diagnostics. Tolland at week two, not yet knowable. BigTime under UT-1; the brief of 21 September. |
| Carve-out pricing | Time and materials for the first four weeks, fixed fee once the separation scope is known. FR-3, 18 September. |
| Open | How Larkspur wants the sprint invoiced. Routed to Marcus; Ellen asks Dan directly and sends what he says to Marcus to record. |
9:52Ellen has four answers with their sources and one question routed to the person who owns it. She decides what goes into the call, and nothing on this page was drafted by software.
No software drafts anything here; the person is the reader.
- What the model does
- Finds the entry that answers the question and quotes it with its source. Approved positions come first, kept records are marked as records, and when neither exists the answer says not recorded. It does not draft, summarize the call, or recommend. That behaviour is written into the instructions file as a rule.
- Where the knowledge lives
- The same memory the six software samples read from: the proposal drafter, the deliverable assembler, the playbook desk, the thought-piece drafter, the Monday brief, and the debrief line. Nothing on this page needed a separate build, which is the whole argument: the memory was already there.
- What gets written back
- The fifth question, routed once to Marcus Delgado, and Ellen's notes for the call. His answer returns as an entry under SP-2 that every later Larkspur Ridge proposal and invoice reads.
- Buildable today
- A Teams app with a question form, and search over the memory scoped to the tenant. Both are inside Microsoft 365 Business Premium, which the firm already pays for.
If Whitlow Ferris Partners never ran any software on this memory, these twenty minutes would still go better than they did, because the firm's positions were written down with their reasoning, an owner, and a date.
App interfaces on this page are simplified illustrations, not vendor screenshots.
The same thing elsewhere
At a law firm this is a partner before a client call, checking what the firm advised that client last time, what it charged, and whether the firm has a position on the fee arrangement being asked for. At an architecture practice it is a principal before a site meeting, checking what the practice found on its last project for that developer and which consultants it used. The workflow changes and the memory underneath it does not.