Hollister Reed LLP: Leadership view

Step 1 of 5: Ground, 14 October 2026

Who are we, what is running, and where did we stand? The written who and why, the dated baseline, and the register of AI with each known owner and every gap.

AI in the business

AI in use: 11 agents and 12 AI toolsfrom the record as found
Vendors supplying AI: 8from the record as found
Serve a written goal: 4 of 23from the record as found
Serve no written goal: 19from the record as found

Who is accountable

Open gaps with a named owner: 43 of 60from the record as found
Governance level: Absent (1.2 of 5)from the record as found
Judgments checked: 0 of 25 scoresfrom the record as found

How much the record can stand behind

Checked against a second source: 0%from the record as found
Entries in the record: 0 of 300 checkedfrom the record as found

Not yet checked: 67 on what was observed, 159 on documents, 74 on what people said.

Value against the baseline

Baseline frozen: 8 October 2026 and 9 October 2026from the record as found
Change since baseline: not yet measured againfrom the record as found
Drivers the sponsor signed: 0 of 5from the record as found

Every result is change since baseline. The twin makes no claim about what caused it.

Decisions and what is still open

Decisions leadership owes: 0from the record as found
Open gaps: 60, of which 17 with no ownerfrom the record as found
Questions asked of the firm still unanswered: 3 of 4from the record as found

No decision is recorded as owed by leadership at this step.

AI economics by use case

No use case has an economics reading yet at this step. Spend is known only by vendor, from the ledger.

Step 2 of 5: Align, 11 November 2026

Which of our AI serves a goal we wrote down, and which serves none? Every entry tied to a written goal, and the gaps named where no line can be drawn.

AI in the business

AI in use: 11 agents and 12 AI toolssimulated
Vendors supplying AI: 8simulated
Serve a written goal: 13 of 23simulated
Serve no written goal: 10simulated

Who is accountable

Open gaps with a named owner: 39 of 56simulated
Governance level: Absent (1.2 of 5)simulated
Judgments checked: 0 of 25 scoressimulated

How much the record can stand behind

Checked against a second source: 5%simulated
Entries in the record: 15 of 296 checkedsimulated

Not yet checked: 63 on what was observed, 159 on documents, 59 on what people said.

Value against the baseline

Baseline frozen: 8 October 2026 and 9 October 2026simulated
Change since baseline: not yet measured againsimulated
Drivers the sponsor signed: 0 of 5simulated

Every result is change since baseline. The twin makes no claim about what caused it.

Decisions and what is still open

Approve or retire the four AI subscriptions paid on company cards that serve no written goal. Owner: Elaine Marsh.

Decisions leadership owes: 1simulated
Open gaps: 56, of which 17 with no ownersimulated
Questions asked of the firm still unanswered: 3 of 4simulated

AI economics by use case

No use case has an economics reading yet at this step. Spend is known only by vendor, from the ledger.

Step 3 of 5: Plan, 9 December 2026

Who closes each gap, by when, and under which controls? A dated plan with an owner and controls for every action, and a day-ninety projection with its assumptions.

AI in the business

AI in use: 9 agents and 12 AI toolssimulated
Vendors supplying AI: 8simulated
Serve a written goal: 13 of 21simulated
Serve no written goal: 8simulated

Who is accountable

Open gaps with a named owner: 50 of 50simulated
Governance level: Absent (1.2 of 5)simulated
Judgments checked: 0 of 25 scoressimulated

The plan's target for 12 January 2027, projected: 2.1, Declared; not yet measured (18 of 25 levels are projected and 7 stand as judged). A projection is a target with its assumptions, not a measurement.

How much the record can stand behind

Checked against a second source: 25%simulated
Entries in the record: 72 of 292 checkedsimulated

Not yet checked: 57 on what was observed, 136 on documents, 27 on what people said.

Value against the baseline

Baseline frozen: 8 October 2026 and 9 October 2026simulated
Change since baseline: not yet measured againsimulated
Drivers the sponsor signed: 0 of 5simulated

Every result is change since baseline. The twin makes no claim about what caused it.

Decisions and what is still open

Approve or retire the four AI subscriptions paid on company cards that serve no written goal. Owner: Elaine Marsh.
Fund firm-approved AI accounts for the tax practice before the busy season. Owner: Elaine Marsh.

Decisions leadership owes: 2simulated
Open gaps: 50, of which 0 with no ownersimulated
Questions asked of the firm still unanswered: 1 of 4simulated

AI economics by use case

No use case has an economics reading yet at this step. Spend is known only by vendor, from the ledger.

Step 4 of 5: Build, 10 March 2027

Is the plan being carried out? The plan's actions closed and measured against the baseline frozen before they began.

AI in the business

AI in use: 9 agents and 8 AI toolssimulated
Vendors supplying AI: 6simulated
Serve a written goal: 13 of 17simulated
Serve no written goal: 4simulated

Who is accountable

Open gaps with a named owner: 32 of 32simulated
Governance level: Declared (2.6 of 5)simulated
Judgments checked: 0 of 25 scoressimulated

How much the record can stand behind

Checked against a second source: 52%simulated
Entries in the record: 144 of 278 checkedsimulated

Not yet checked: 33 on what was observed, 75 on documents, 26 on what people said.

Value against the baseline

Baseline frozen: 8 October 2026 and 9 October 2026simulated
Change since baseline: 6 of 10 measures changed since baselinesimulated
Drivers the sponsor signed: 1 of 5simulated

Gross margin: +0.8 percentage points since baseline.

Revenue per employee: +$8,021 since baseline.

Audit report cycle time: -1.1 days since baseline.

Client onboarding cycle time: -0.8 days since baseline.

Return preparation cycle time: -0.8 days since baseline.

Client churn: -0.5 percentage points since baseline.

Every result is change since baseline. The twin makes no claim about what caused it.

Decisions and what is still open

Accept the residual risk from client documents uploaded before the matrix, or commission a review of them. Owner: Elaine Marsh.

Decisions leadership owes: 1simulated
Open gaps: 32, of which 0 with no ownersimulated
Questions asked of the firm still unanswered: 0 of 4simulated

AI economics by use case

Every figure in this table is simulated.

Use case and owner	AI spend, twelve months	Tokens, latest month	Cost per task	Hard gain, twelve months	Soft gain, latest month	Quality	Attribution
Copilot pilotLuis Herrera	$1,800	not reported	not measured	none recorded	40 hours a month (estimate $3,800)	not measured	not attributed
Email and memo draftingGraham Whitlock	$450	1.5 million in February 2027	$0.36 in February 2027	none recorded	20 hours a month (estimate $1,900)	not measured	not attributed
Prior-year return summarizationRuth Adeyemi	$270	1.1 million in February 2027	$0.30 in February 2027	none recorded	25 hours a month (estimate $2,375)	not measured	not attributed
Proposal draftingNora Castillo	$540	2.3 million in February 2027	$4.86 in February 2027	none recorded	30 hours a month (estimate $2,850)	not measured	not attributed
Returning-client briefingA tax partner	$660	3.2 million in February 2027	$1.55 in February 2027	none recorded	45 hours a month (estimate $4,275)	not measured	not attributed
Routine client message answeringPeter Lindqvist	$420	2.0 million in February 2027	$0.16 in February 2027	none recorded	60 hours a month (estimate $5,700)	not measured	not attributed

AI spend, twelve months: $4,140simulated
Of it, with no named owner: $0simulated
Of it, serving no written goal: $420simulated
Hard gains, attributed: $0simulated
Hard gains, not attributed: $0simulated
Soft gains, last month: 220 hours a month (estimate $20,900)simulated

Each figure is a dated reading from the ledger, the vendors' monthly usage reports, the systems' task logs and the users' monthly time surveys, taken after the period it covers.

Hard gains appear in the ledger, as revenue billed or cost removed. Soft gains are hours people report saving, with a dollar estimate at the firm's stated loaded rate. The two are never added together.

A gain counts as attributed only when the use case's owner has signed for it; every other gain is shown as not attributed.

Step 5 of 5: Keep, 14 October 2027

Does the map stay true as AI changes? The Twin, licensed and running in the firm's own environment, measured again against the baseline on the occasions the firm chooses.

AI in the business

AI in use: 9 agents and 9 AI toolssimulated
Vendors supplying AI: 7simulated
Serve a written goal: 13 of 18simulated
Serve no written goal: 5simulated

Who is accountable

Open gaps with a named owner: 34 of 35simulated
Governance level: Reported (3.1 of 5)simulated
Judgments checked: 12 of 25 scoressimulated

How much the record can stand behind

Checked against a second source: 72%simulated
Entries in the record: 206 of 288 checkedsimulated

Not yet checked: 25 on what was observed, 36 on documents, 21 on what people said.

Value against the baseline

Baseline frozen: 8 October 2026 and 9 October 2026simulated
Change since baseline: 6 of 10 measures changed since baselinesimulated
Drivers the sponsor signed: 2 of 5simulated

Gross margin: +0.8 percentage points since baseline.

Revenue per employee: +$8,021 since baseline.

Audit report cycle time: -1.1 days since baseline.

Client onboarding cycle time: -0.8 days since baseline.

Return preparation cycle time: -0.5 days since baseline.

Client churn: -0.5 percentage points since baseline.

Every result is change since baseline. The twin makes no claim about what caused it.

Decisions and what is still open

Fund an independent test of the Copilot controls before the next insurance renewal. Owner: Elaine Marsh.
Accept the residual risk from client documents uploaded before the matrix, or commission a review of them. Owner: Elaine Marsh.
Approve or retire the note-taking service found in the September 2027 ledger, and name its owner. Owner: Elaine Marsh.

Decisions leadership owes: 3simulated
Open gaps: 35, of which 1 with no ownersimulated
Questions asked of the firm still unanswered: 0 of 4simulated

AI economics by use case

Every figure in this table is simulated.

Use case and owner	AI spend, twelve months	Tokens, latest month	Cost per task	Hard gain, twelve months	Soft gain, latest month	Quality	Attribution
Copilot pilotLuis Herrera	$13,200	not reported	not measured	$9,400 cost removed	110 hours a month (estimate $10,450)	not measured	not attributed
Email and memo draftingGraham Whitlock	$1,500	1.8 million in September 2027	$0.33 in September 2027	none recorded	22 hours a month (estimate $2,090)	not measured	not attributed
Prior-year return summarizationRuth Adeyemi	$1,080	1.6 million in September 2027	$0.34 in September 2027	none recorded	28 hours a month (estimate $2,660)	17.5% corrected, 2.8% overridden	not attributed
Proposal draftingNora Castillo	$2,160	3.1 million in September 2027	$4.53 in September 2027	$89,300 revenue billed	34 hours a month (estimate $3,230)	6.3% corrected, 1.05% overridden	signed by Nora Castillo
Returning-client briefingA tax partner	$2,440	3.7 million in September 2027	$1.64 in September 2027	none recorded	52 hours a month (estimate $4,940)	8.4% corrected, 2.1% overridden	signed by A tax partner
Routine client message answeringPeter Lindqvist	$1,520	2.4 million in September 2027	$0.17 in September 2027	none recorded	64 hours a month (estimate $6,080)	4.89% corrected, 1.75% overridden	not attributed

AI spend, twelve months: $21,900simulated
Of it, with no named owner: $0simulated
Of it, serving no written goal: $1,520simulated
Hard gains, attributed: $89,300simulated
Hard gains, not attributed: $9,400simulated
Soft gains, last month: 310 hours a month (estimate $29,450)simulated

Each figure is a dated reading from the ledger, the vendors' monthly usage reports, the systems' task logs and the users' monthly time surveys, taken after the period it covers.

Hard gains appear in the ledger, as revenue billed or cost removed. Soft gains are hours people report saving, with a dollar estimate at the firm's stated loaded rate. The two are never added together.

A gain counts as attributed only when the use case's owner has signed for it; every other gain is shown as not attributed.