AI Blueprint · 25 July 2026Sample - fictional firm
Making Marlow & Finch AI‑first, by design.
This Blueprint shows where to start, how ~$569k and ~2,280 hours a year of current delivery time can be handed back, what additional revenue may become possible, and the controls that keep your firm's judgement in charge. Document production is the thread through it all.
Findings, priorities and 12-month roadmap
Marlow & Finch Lawyers is a fictional firm. The method, scoring and structure are exactly what we produce for a real engagement.
Prepared for Priya Raman, Tom Whitfield, Alice Ng and the Marlow & Finch team.
Blueprint workshop · Sunshine Coast · 22 July 2026
Prepared by Bart Puszko · Blue Seas AI
Queensland Government AI Mentor2025 Australian AI Awards Finalist2025 SC Business Award - Advanced Tech
Executive summary
Start with matter documents. They give your firm the fastest proof.
Marlow & Finch is not starting from scratch. The precedents, matter plans, costs agreements and review habits already describe good work. The constraint is the fee-earner time spent turning a file into finished documents. A governed first draft gives that time back while your solicitors keep every judgement and approval.
~$569k / year
Current mapped charge-out cost-to-serve handed back across intake, matter delivery and document production.
ESTIMATE
~2,280 hrs / year
Fee-earner and support time handed back across the same current-workflow models.
ESTIMATE
1 first build
Matter Delivery Engine - the clearest controlled proof that drafting time can fall without quality falling.
RECOMMENDATION
What we found
Document production is the repeated bottleneck. A property matter carries about 6.4 hours of file work, while about five advices and client updates a week move through one senior solicitor.
Why start there
The inputs and output are structured, the firm has strong precedents, and Alice Ng can judge a draft quickly. That makes the first test narrow and measurable.
What your firm gets
A validated file becomes a Marlow & Finch formatted first draft with every statement traced to its source. Alice Ng keeps the legal calls and document approval. The responsible solicitor keeps the advice.
What returned time may enable
$168k-$224k from 60-80 additional property matters and $288k from four additional retainer clients are separate conditional revenue scenarios, held outside the ~$569k.
The Blueprint answer. Build one matter-document proof, test whether it saves more editing than it creates, then extend the proven pattern to advice production and client intake. Stop if traceability, quality or the time-saving gate fails.
The ~$569k is recoverable charge-out cost-to-serve. It is an estimate. It excludes revenue, cash, margin and any guarantee, and it excludes the alternative $37,200 intake internal-cost view, conditional property and retainer revenue, overlapping foundation value and TBC partner-advisory economics.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Scope, method and evidence
How to trust the recommendation and the numbers.
Every workflow step shows what AI can safely do, what must stay human and how firm the evidence is. That keeps the recommendation useful without pretending an estimate is a fact.
The two scores on every step
S
AI Safety /10
How safe AI's proposed contribution is under the named controls. Reviewed, internal or reversible work can score high even when your firm keeps the decision human.
C
AI Capability /10
How well today's models perform the AI-shaped work. A retained relationship, judgement or approval does not lower the score; the map colour records the firm's operating choice.
$
Value handed back
The model returns 100% of a fully automated step, 75% of an AI-assisted step and 0% where the work stays human. Charge-out rates value the time used to deliver the work.
The assumptions behind every dollar
Fee-earner time
Support time
Recovery rule
Valued at the firm's realised charge-out rate of $480/hr across 9 solicitor fee earners.
Valued at loaded cost of $62/hr for practice support, paralegal and administrative time.
100% of a fully automated step, 75% of an AI-assisted step, 0% where the work stays human.
Scope and method limits
In scope
Out of scope
Method limit
Client intake and engagement, matter delivery, advice and document production, partner-advisory boundaries, the Workbench and the 12-month sequence.
Trust account movements, court filings, production security validation, contract design and unapproved client material.
Intake, matter delivery and document-production stage allocations are estimates anchored to the firm's aggregate effort. Validate them before pricing or workload commitments.
How firm is each item?
ObservedSeen in an artefact or workflow.
StatedProvided by the firm.
AssessedBlue Seas AI judgement.
EstimateCalculated from stated inputs.
TBCInput still required.
Confidential working document. Share within the firm, Blue Seas AI and approved advisers.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Gaps and opportunities
Document production is the bottleneck your firm can fix first.
Six workshop findings explain where fee-earner time goes, why growth stalls and why a governed first draft is the practical starting point.
1Document production is the firm-wide bottleneck
Stated · artefact reviewed
A property matter follows a repeatable structure, yet the file work still takes about 6.4 hours. Advices and client letters carry the same drafting drag every week.
“You should be able to put the file in and have the first draft come back to you the same morning.”- Priya Raman, Managing Partner
2Partners are tied up in delivery
Stated
The firm can take on more matters once fee-earner time moves out of marking up drafts and back into acting on the work and speaking to clients.
“We cannot take on more because we are flat out delivering what we already have, and that is where the week goes.”- Priya Raman, Managing Partner
3Advice production rides on one senior solicitor
Observed
About five advices and client updates a week move through Alice Ng. Collection and repeated sections suit a governed first draft. The legal position stays with the solicitor.
4Intake loses time before drafting begins
Estimate · stated input
About eight new enquiries arrive each week. Opening one matter takes about 2.6 hours and $579 of firm time before a fee earner starts the legal work, and every act-or-decline call waits for Priya Raman.
5Matter delivery and advice depend on key people
Observed
Property matters lean on Alice Ng and advice work on the same small group. Their expertise already lives partly in precedents. Reusable first drafts let juniors get further before a senior needs to step in.
6The build has strong raw material
Observed · artefact reviewed
The matter plans, precedent set, checklists, house document structures and costs agreements describe the work. The build can start from the firm's own standards.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
What we need from you
Three inputs turn this recommendation into a build.
Marlow & Finch does not need to prepare a perfect data room. These inputs let Blue Seas AI confirm the baseline, define the test and return a separate fixed-scope proposal.
1
Access and introductions
Show us where the working material lives across Microsoft 365, SharePoint, the document management system and your practice management system. Tom Whitfield can then help scope the proposed Australian-hosted AI environment, data boundaries and approval points. Initial testing uses closed or redacted matters.
Priya Raman + Tom Whitfield
2
The few artefacts we have not seen
The matter plan, file-opening checklists, one template engagement letter, one completed advice and an export of the precedent library. These give the first build your firm's structure, language and review standard.
Tom Whitfield coordinates · Alice Ng supplies examples
3
The numbers that firm up the estimates
Enquiry volume and conversion rate, conflict-check effort, current retainer client count and observed stage timings where the workshop sheet was blank. Ranges are enough. Each estimate becomes a measured baseline before it becomes a build target.
Priya Raman + process owners
What happens next
Once the firm chooses the first play, Blue Seas AI returns with a scoped Phase 1 proposal. The proposal defines the data boundary, test set, human approvals, success measure and stop condition before any live client material is used.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Operating map and companions
Open any workflow to see where the time goes.
The live maps show each step, who owns it, what it costs, what AI may handle and where a person still decides. Partner-led advisory remains a separate human-led service.
Cost-to-serve is the opportunity value of delivery effort. It excludes your client price, gross margin, cash saving and the Blue Seas AI implementation fee.
GradesAll map dollars are estimates. Exact quotes are verified against the workshop record; modelled stage splits are labelled Estimate and stay subject to your validation.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Margin and value overview
Current work can hand back ~$569k and ~2,280 hours.
The bars value delivery effort your firm already spends. The three cards show separate possibilities: time value in the current practice and revenue only if the firm wins and delivers additional work.
Client intake and engagement
modelled path per new matter · 2.6 hrs of effort · about 400 new matters a year · per-matter dollars are rounded, the annual total is calculated from hours · conversion rate TBC
Recoverable $344
Human / retained $235
Cost-to-serve $579 per matterAnnual charge-out workload $231,600Hours returned~600 hrs/year
Matter delivery
modelled property path · 6.4 hrs per matter · 25% fee earner / 75% support · about 255 matters a year; validate before pricing
Recoverable $666
Human / retained $400
Cost-to-serve $1,066 per matterAnnual charge-out workload $271,830Hours returned~1,020 hrs/year
Advice and document production
about 5 advices and client updates a week, roughly 264 a year · 4 hrs each · 62.5% recovery leaves 1.5 hrs of solicitor time on every document
Recoverable $991
Solicitor / retained $595
Cost-to-serve $1,586 per documentAnnual charge-out workload $418,704Hours returned~660 hrs/year
Intake time already being spent
$37,200
Modelled annual value of the same 600 intake hours at the $62/hr loaded internal cost.
Estimate · time value
If the firm takes 60-80 more property matters
$168k-$224k
Conditional gross revenue at the stated $2,800 fixed fee, subject to demand and delivery time.
Scenario · revenue
If the firm adds four retainer clients
$288,000
Conditional gross revenue at the workshop's $6,000 per month scenario.
Scenario · revenue
How to use these numbers. First measure whether the build returns hours from work your firm already performs. Only then decide how to use those returned hours - improve turnaround, protect quality or take on more matters. The blue $37,200 card is an alternative internal-cost view of the same intake time. The green and amber cards are conditional gross-revenue scenarios. All three sit outside the ~$569k current-workflow recovery total.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Plays ranked
Six plays, one decision view.
The first two production systems carry the current annual value. The next four make those systems safer, repeatable and easier for your firm to run. Foundation value is included above or remains TBC, so it is counted once.
#
Play
Journey line
AI Safety
AI Capability
Hands back / year
Verdict
1
Matter and Advice Production Engine
Matter delivery + advice
8
9
~$431,454 + ~1,680 hrs
Build
2
Client Intake and Engagement Factory
Client intake
8
9
~$137,520 + ~600 hrs
Build
3
Quality Evaluator
All 3
8
8
Included above · review time TBC
Build
4
Matter Opening and Handover Runner
Intake + matter delivery
9
8
Included above · stage time TBC
Build
5
Matter File Collator
Matter delivery
8
9
Included above · reuse across matter types
Build
6
Precedent and Knowledge Layer
All 3
7
7
Value TBC · resilience and reuse
Extend
Total current mapped recovery~$569k / year · ~2,280 hours
Client intake $137,520 + matter delivery $169,830 + advice production $261,624 = $568,974. Hours: 600 + 1,020 + 660 = 2,280. Charge-out cost-to-serve only. It excludes revenue, cash and any guarantee.
Value rank is not build order. The roadmap starts with matter delivery because it offers the clearest, safest proof. The standalone roadmap shows both the judgement-led sequence and the economic-only alternative.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
The biggest wins, first
Two systems create value. One evaluator keeps control.
These are the operating systems behind the annual recovery on the previous page. Each starts with a complete first draft and ends with a named human decision.
1
Matter and Advice Production Engine · Matter delivery + advice · AI Safety 8 · AI Capability 9
Your solicitors start with complete, structured drafts
Hands back: ~$431,454 / year + ~1,680 hours across the current matter delivery and advice models
Today. A property matter carries about 6.4 hours of file work before settlement. Advice and document production runs about five advices, letters and client updates each week through a repeatable but manual path.
The proposed build. Two governed recipes turn a validated file and collected precedents into Marlow & Finch formatted first drafts, contract and search summaries and client letters, ready for solicitor review.
What stays human. Alice Ng owns document review and the legal calls. The responsible solicitor owns the advice and the final position. Both approve the finished work.
2
Client Intake and Engagement Factory · Client intake · AI Safety 8 · AI Capability 9
Priya Raman sees the decision and the draft, ready for judgement
Hands back: ~$137,520 / year + ~600 hours across the current intake workload model
Today. The modelled intake path costs $579 per new matter to serve. Annual intake workload is modelled at about 1,040 hours across roughly 400 new matters; enquiry conversion rate remains TBC.
The proposed build. Enquiry triage captures the facts and runs the conflict search, a drafter assembles the costs disclosure and scope into an engagement letter, and an evaluator checks scope, rates and commitments.
What stays human. Act or decline, the conflict clearance, the fee position, the final review and every client commitment.
Your people review exceptions, with the final approval still theirs
Hands back: included in the mapped journey totals; observed review time remains TBC
Today. Senior solicitors read whole drafts for tone, source, costs and commitment errors.
The proposed build. A separate evaluator checks each draft against your firm's rubric and returns an exception report with the source material attached.
What stays human. The final judgement, approval and client commitment.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
The foundation and the flywheel
Three shared capabilities make the first wins repeatable.
These plays reduce rework and key-person dependency across the first systems. Their standalone time value stays TBC because the workshop did not supply complete stage timings.
Play 4 · Intake + matter delivery
Matter Opening and Handover Runner
Safety 9 · Capability 8
Build
Today
The matter plan, file structure and practice management system setup are created by hand at the start of each matter.
The proposed build
Assemble the approved file-opening pack and create the handover in your practice management system from the accepted engagement, with fee material removed.
What stays human
The client conversation, conflict clearance, owner assignments and approval to open the file.
Play 5 · Matter delivery
Matter File Collator
Safety 8 · Capability 9
Build
Today
Contract terms, searches and correspondence are pulled together by hand before any document can be drafted.
The proposed build
Match the file material to the matter checklist, preserve the source of every item and flag gaps for Alice Ng to decide.
What stays human
Document review, the sufficiency call and every legal position.
Play 6 · All 3 journeys
Precedent and Knowledge Layer
Safety 7 · Capability 7
Extend
Today
Property matters lean on Alice Ng and advice work on the same small group. The firm is already building the precedents that reduce that dependency.
The proposed build
Turn approved precedents and worked examples into the one governed reference each agent and junior solicitor can use.
What stays human
The senior solicitors set the standard, approve changes and decide when an exception needs them.
Why these come next
Start small, build trust, then build bigger. The first two systems prove the value. These three make the operating model easier for your firm to own.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Process scorecard and maturity
The work is documented. Controlled automation is next.
The matter plans, precedents, house document structures and costs agreements mean your firm already owns the process. The work now is putting governed automation around standards your people already own.
Matter delivery conveyancing and property
Ready to automate
1People-dependent
2Documented
3Standardised
4AI-assisted at scale
Advice and documents advices and client letters
Ready to automate
1People-dependent
2Documented
3Standardised
4AI-assisted at scale
Client intake enquiry to engagement letter
Standardise, then automate
1People-dependent
2Documented
3Standardised
4AI-assisted at scale
The story in one line
Marlow & Finch has already systematised the work. Automate the repeatable parts first, starting with matter delivery and advice production where the precedents and examples are strongest.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
The Marlow & Finch AI Workbench
Nine small agents, one governed practice system.
The proposed design uses nine small agents. Each has one defined task, reuses approved firm knowledge and stops where a named person must decide.
How to explore it. Open the interactive Workbench, choose a practice workflow, then click each agent to see the task it will be built to handle, the systems it will use and who still approves the result.
Routes work and enforces gates
Practice Operations Orchestrator
Added only after the first agents pass their quality and safety tests. It routes approved work; the people who own the decision keep it.
Enquiry Triage
Automate
Captures new enquiries, extracts the matter facts and prepares the act-or-decline brief.
Engagement Letter Drafter
Review
Builds scope and costs disclosure into a tailored engagement letter draft.
Matter File Collator
Review
Collates contract terms, searches and correspondence and flags gaps for a solicitor.
Advice Builder
Review
Builds Marlow & Finch formatted advices, letters and file notes.
Precedent Collector
Automate
Collects, de-duplicates and enriches approved precedents and source material.
Client Update Drafter
Review
Drafts repeated client update sections; the responsible solicitor owns the advice.
Evaluator
Hard gate
Checks sources, completeness, tone and approved criteria.
Librarian
Automate
Retrieves approved precedents, worked examples and prior matters.
Matter Runner
Automate
Opens the matter, creates the file structure and handover from approved work.
Proposed controlled boundary: Microsoft 365 and SharePoint for approved knowledge · your practice management system for matter state · proposed Australian-hosted model endpoint · versioned evaluation and audit records before production.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Partner-led advisory
Advice stays human. AI prepares the file.
Partner-led advisory was not costed because the workshop did not provide a mapped volume or effort baseline. The useful decision is the boundary: where AI prepares and where a trusted adviser must remain in the room.
Human owns
Client trust, education and the in-room relationship.
The legal risk call, the trade-off and the final recommendation.
Negotiation, court appearances and every undertaking given.
Trust account movements and court filings, which stay human by design.
Approval of every client-facing advice or strategy.
AI supports
Retrieves approved firm precedents and prior matters.
Drafts advice, letter, file note and meeting scaffolds.
Builds briefing packs, action registers and matter records.
Flags missing material on the file and internal inconsistencies.
Helps junior solicitors prepare without interrupting a partner for every step.
Evidence basis“What the client needs is a person across the table who can explain it, look them in the eye and talk them through the choice.” - Priya Raman, Managing Partner The workshop also exposed a knowledge dependency where a junior solicitor needed a partner to step through the settlement risk position.
No economic claimAdvisory volumes, hours, rates and return are all TBC. Trust account and court-filing work is excluded from the economics because it stays human by design.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
12-month roadmap
Start where proof is clearest, then extend what works.
Matter delivery comes first because its artefacts, pain and output are easiest to test safely. Advice production follows the same drafting pattern. Client intake comes after the production proof, and the Workbench is assembled only when its workers have passed their gates. Timing is directional and often faster in practice; every phase remains gate-dependent. A failed pause condition means correct and retest before progressing; the programme continues.
Phase 1
Approx. months 1-3
Matter Delivery Engine
Proposed: Alice Ng process · Bart delivery · Priya Raman + Tom Whitfield approve · joint gate Build from validated contract terms, searches and the house document set, with the legal calls made by a solicitor. Exit: agreed eval set, traceable draft and two human QA stages. Pause / correct: any material term, key date or recommendation lacks a source on the file, or any material error survives final human QA.
Phase 2
Approx. months 4-6
Advice and Document Production
Proposed: Alice Ng process · Bart delivery · Priya Raman + Tom Whitfield approve · joint gate Scoped at gate. Preserve the source of every statement and solicitor ownership of the advice. Pause / correct: client matters cannot be reliably separated, sources are missing or solicitor review time does not improve across the pilot outputs.
Phase 3
Approx. months 7-9
Client Intake and Engagement
Proposed: Tom Whitfield process · Bart delivery · Priya Raman approves · joint gate Scoped at gate. Add enquiry capture, conflict search support, engagement letter drafting and evaluator checks. Pause / correct: any scope, costs or commitment error survives review, a conflict match is mishandled or preparation time does not improve across the pilot outputs.
Phase 4
Approx. months 10-12
Governed Workbench + Become AI‑First
Proposed: Tom Whitfield operations · Bart delivery · Priya Raman + Tom Whitfield approve · joint gate Add the orchestrator only after workers and gates are proven. Build the firm's internal capability while Blue Seas AI builds the practice systems. Pause / redesign: any matter-separation, approval, audit or rollback test fails, or no named owner at the firm can operate the workflow after supervised use.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Guardrails and data flow
Trust account movements and court filings stay human by design.
That is the line the whole design is built around. Client confidentiality and legal professional privilege carry the same weight, so the build honours the standard you already hold yourself to under the Australian Solicitors Conduct Rules and current Australian guidance on securing AI agents.
Trust and the court stay human
No agent initiates, authorises or records a trust account movement. No agent lodges a court document or sets a filing deadline. Those steps stay with a named person who holds the practising certificate and the authority, with no system access granted to any agent.
Australian‑hosted and matter‑scoped
Proposed hosting on an Australian-hosted AI environment inside your firm's tenant. Each client's matter is separated with its own access and context. Australian data location, subprocessors and retention must be approved and tested before live use.
Human approval on decision steps
Controlled context, an instruction hierarchy, distinct agent identities and layered checks. The legal calls, the advice and final document approval stay with named solicitors.
Verified deliverables
Every AI-assisted document is checked before it leaves the firm. Each statement retains its source and the operating standard can align with current Australian AI guidance without implying a certification the firm does not hold.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Guardrails and data flow
Six controls must be approved before live client material.
Each control names the risk, the operating response and the person or cadence that owns it. Every one is a working gate for the build.
Trust account movements and court filings stay human
ID
Risk
Required control
Owner / cadence
AI-01
Confidentiality or privilege leaks between matters
Matter-scoped stores, context labels, separate access policy and a contamination test each release.
Tom Whitfield + Blue Seas AI · each release
AI-02
Conflict cleared on incomplete information
Conflict search runs against the full client and related-party register; an unresolved match stops the file from opening.
Alice Ng · every new matter
AI-03
Advice or correspondence unsupported by the file
A source on every generated statement, approved material only, an instruction hierarchy that ignores directions embedded in client documents, and named solicitor approval.
Responsible solicitor · every output
AI-04
Limitation date, undertaking or filing deadline missed
Key dates and undertakings are extracted for a person to confirm; a named solicitor makes and checks the diary entry, and no agent lodges or files.
Alice Ng · every matter
AI-05
A change degrades quality after go-live
Versioned tests, rollback, change approval and monthly assurance evidence.
Marlow & Finch + Blue Seas AI · each change
AI-06
Data location or retention mismatch
Approved Australian endpoint, documented subprocessors, retention and deletion tests, and a review confirming no agent holds trust account or lodgement access.
Tom Whitfield · quarterly
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
Where to from here
Start with one controlled proof: the Matter Delivery Engine.
The Blueprint makes the decision visible. Nothing is built until Marlow & Finch accepts a separate proposal. Each later phase is earned by evidence, security and a measured time-saving result.
1
Your AI Blueprint
Your firm now has the findings, mapped economics, ranked plays, guardrails and roadmap. The immediate decision is whether matter documents are the right first proof.
2
Accept a separate build proposal
Blue Seas AI returns a fixed-scope Phase 1 proposal covering data, approvals, security, test cases, success measures and the stop condition. Work starts only after acceptance and may run over several months through measured gates.
3
Learn AI by doing - Become AI‑First
Blue Seas AI works alongside your team on real matters so named owners finish ready to run, trust and extend what has been built.
What proof looks like
Alice Ng receives a cleaner Marlow & Finch formatted starting point, every statement still traces to the file and the draft saves more editing time than it creates.
When the test stops
Stop if source traceability fails, material errors exceed the agreed threshold or human review takes as long as drafting the document from scratch.
“Look at all of it, then make us AI by design. We build our files that way and we should build this the same way.”- Priya Raman, Managing Partner
The decision now
If the Matter Delivery Engine is the right first move, confirm that play. Blue Seas AI will return with the scoped Phase 1 proposal and take it one gate at a time.
Marlow & Finch Lawyers is a fictional firm and this is a sample. Figures are productivity and separate conditional revenue scenarios based on firm-supplied inputs, and they are estimates. Implementation is separately scoped and contracted.
Marlow & Finch - AI Blueprint · Sample · We make your business AI-first.
