For Rachel Sumner · Morr & Co

Everything we talked about, in one place.

The agent for Dennis's Spanish series. A thirty-six agent matrix so you can see the other ways AI can help your marketing team. The synthetic users approach. The GEO audit prompt. The Frontier Program brief for Greg. Pick what's useful. Ignore the rest for now.

The House Style Editor for Dennis's Spanish series.

You told me you'd spent two months going backwards and forwards on Dennis's CrossBorder articles, trying to bridge his voice with the firm's. The agent is built. It takes Dennis's raw draft and returns a Morr & Co version. Most importantly, it ships with a Legal Substance Audit so Dennis can see, line by line, that nothing legal was changed in the rewrite.

Built and in your inbox

How to run it

The five agent files are attached to the email I sent. Drop them into Microsoft Copilot, attach Dennis's raw draft, and run. Full step-by-step instructions and screenshots are on our how-to page:

How to use your agent →

Safety story for Mike

If Mike asks how we're confident the agent won't change the law, here's the one-line answer: before any rewriting, the agent extracts every legal claim, statute reference, jurisdictional fact, procedural step and case-study fact from Dennis's draft into a locked list. Every item on that list must appear in the rewritten article with the same meaning, direction and force. The audit table proves it, line by line. A single failure in the audit and the agent refuses to deliver. Dennis sense-checks the audit before publication, so the law is always in his hands.

Thirty-six AI agents for your role.

The House Style Editor is one of thirty-six AI agents we could build for the kind of work you do. The matrix below maps six of your KPIs (rows) against the six AI mechanisms (columns). Each cell is a different agent. Don't try to pick six. Find the one that makes you exhale, and start there. One is already on your desk.

KPI ↓
Mechanism →
Column 1
Compress
Column 2
Amplify
Column 3
Predict
Column 4
Personalise
Column 5
Inspect
Column 6
Decide
Row 1
Content production time
The Partner Briefing Stenographer
Partner records a five-minute voice note; you get the article draft, the LinkedIn post and the newsletter teaser.
−3 hrs per article
The Single-Source Multiplier
One approved article spawns the LinkedIn long-form, the carousel, the email teaser and the print snippet, in house voice.
4x channel coverage
The Calendar Stress Forecast
Looks at the editorial calendar four weeks out and flags the weeks that will overload your three-person team before they happen.
−1 weekly fire-drill
The Channel Adapter
Same article, reshaped per channel: long-form for LinkedIn, snippet for email, headline for print quarterly. One pass.
−45 min per piece
The Drafts Triage
Every partner submission gets graded for usefulness, novelty against existing coverage, and ROI fit, before you start work on it.
−2 dead-end articles per month
The Editorial Queue Optimiser
Given eight partner submissions and four publishing slots, picks the four with the highest predicted reach for your audiences.
+30% throughput
Row 2
Brand voice across ten practice areas
Built · In your inbox
The House Style Editor
Partner sends a raw draft; you get a Morr & Co version, with a legal-substance audit so nothing changes that shouldn't. Already built for Dennis's series.
−2 months per series
The Voice Replicator
One approved article becomes five reusable templates the partner fills in next time. Rewrite cost drops with every cycle.
−50% rewrite over six articles
The Tone Drift Forecaster
Reads the last twenty articles and predicts which partners are most likely to drift off-voice in the next quarter.
Heads-up at quarter start
The Partner-Voice Calibrator
Each partner has a signature file; the agent honours their voice while keeping firm tone consistent. Both sound right.
Higher partner adoption
The Site Voice Sweep
Crawls the website monthly and flags pages where the voice has drifted from house style.
Quarterly cleanup, not annual
The Tone Arbiter
When two partners disagree on phrasing, the agent shows house precedents and recommends, with citations.
−1 escalation to Mike per month
Row 3
Content performance ROI
The HubSpot Story
Compresses last month's HubSpot analytics into a three-paragraph "what worked, what didn't, what to test next" brief for Mike.
Monthly brief in 5 min
The High-Performer Multiplier
Takes the top-performing article from the last quarter and produces variants for adjacent practice areas.
3 to 5 new articles from 1 win
The Click-Through Predictor
Before publication, predicts which articles will pull and which won't, with the diagnosis attached.
−30% wasted pages
The Reader-Segment Reframer
Same article rewritten for different reader types: private client, corporate, in-house counsel. Choose where to send.
+2x relevance
The Underperformer Diagnostic
Reads underperforming articles alongside their HubSpot data and explains what specifically caused the drop.
Targeted fixes, not guesses
The Cut-or-Keep Auditor
Looks at the website's article library and recommends update, retire, or keep, with reasoning.
Smaller, sharper library
Row 4
Newsletter and channel coordination
The Monthly Compiler
Takes eight partner submissions and assembles the corporate insights newsletter, HubSpot-ready, in your voice.
−4 hrs per issue
The Newsletter Echo
One newsletter produces five LinkedIn snippets across the following month, so the content keeps echoing.
5x reach per issue
The Send-Time Optimiser
Predicts the best send-time per newsletter based on past open-rate patterns and recipient timezone clusters.
+10% open rate
The Reader-Personalised Newsletter
Each recipient gets a slightly different newsletter, prioritising articles their practice-area history suggests they care about.
+20% click-through
The Newsletter Health Check
Audits each issue's subject lines, preheaders and layout against industry benchmarks before send.
−1 missed mistake per issue
The Print-vs-Online Allocator
Decides which quarterly content goes to print and which stays online, based on shelf-life and audience fit.
Cleaner print issue
Row 5
Website, SEO and GEO health
The Site Audit Summariser
Compresses an SEO and GEO crawl into a one-page partner-friendly summary of the top five things to fix.
Partner-ready in 1 page
The FAQ Generator
Every existing service page spawns ten AI-friendly FAQs in house voice. Direct GEO uplift, no rewrite needed.
+10 FAQs per page
The Search Drift Forecast
Predicts which pages will lose traffic in the next three months based on AI search behaviour patterns.
Early warning, not autopsy
The Practice-Area Persona Page
Each of the ten practice areas gets a landing page tuned to its archetypal reader. Same source content, ten shapes.
+25% page time
The GEO Auditor
Runs the website through the GEO checklist: questions as headings, FAQs present, conversational language, schema. Reports gaps.
Full audit in a day
The Page Triage
Looks at all service pages and recommends the rewrite order: next five, expected lift each.
Roadmap in 1 hour
Row 6
Comms strategy alignment
The Stakeholder Briefer
Compresses Catherine's three-year plan into one-page weekly briefs, one per department head, in their language.
−2 hrs per week briefing
The Strategy Echo
One quarterly strategy doc spawns the all-firm comms, the partner brief, the internal newsletter and the external statement.
4 outputs from 1 doc
The Capacity Forecast
Predicts when the three-person team will hit overload given the next six months of partner-driven asks.
Conversation with Mike, not crisis
The Department-Tailored Plan
One comms strategy with ten personalised slices, one per practice area, so every partner sees themselves in it.
Partner buy-in by design
The Strategy Gap Auditor
Compares the three-year plan against this quarter's actual activity and flags where activity doesn't match strategy.
Quarterly course-correct
The Roadmap Sequencer
Given twenty strategy initiatives and limited capacity, recommends the sequence and what can run in parallel.
Confident roadmap

What the six mechanisms mean

1. Compress

Take a lot of input and distil to a useful little. Hours of work into minutes of reading.

2. Amplify

Take one input and multiply it across channels, audiences or formats. One piece into ten.

3. Predict

Forecast what's likely before it happens. Retrospective dashboards become early-warning systems.

4. Personalise

Tailor the same underlying content to different readers, partners or practice areas. Same source, ten shapes.

5. Inspect

Find the issues a tired pair of human eyes will miss. Audits, drift checks, gap analyses.

6. Decide

Take a list of options and recommend the best one, with reasoning. Replaces the "where do I start" pause.

Test your comms before they leave the office.

AI-generated personas, calibrated to behave like your real audience, that you can run a message past in minutes and get a strong read on how the real audience would respond. Used by The Times and large consumer brands. Particularly useful when you're testing headlines, standfirsts, subject lines, or the question "is this article worth running at all".

Guide and prompts

The full Synthetic Users page

Open the guide on the AI Potential proposals site for the panel construction approach, the prompts, and the calibration loop.

Open the Synthetic Users page →

The website audit prompt I promised you.

The webinar you watched on AI brand audit was right that you can do most of this yourself. Here's the prompt to drop into Copilot. It works on any single page; for the full site you run it on each major section in turn.

Prompt

The full prompt

Act as an expert in Generative Engine Optimisation (GEO). I want you to audit the following web page for how well it will perform in AI search results (ChatGPT, Claude, Perplexity, Copilot, Gemini, Google's AI Overview). Page URL or pasted content: [paste URL or content here] Audience and intent: [one line on who reads this and what they want, e.g. "UK-based business owners considering selling a Spanish property, looking for plain-English guidance"] For your audit, score the page out of 10 on each of the following GEO factors. For each, give a short reason and one concrete fix: 1. Questions as headings (does the page use real reader questions as H2/H3 headings, e.g. "What is the legal framework for...") 2. FAQs present (does the page include an FAQ block of at least 5 question-answer pairs) 3. Conversational language (does the prose read like a person explaining, not like a brochure) 4. Direct, scannable answers (does each section open with a one-sentence answer before the explanation) 5. Citations and proof points (does the page reference real sources, dates, regulations, named experts) 6. Schema and structure (does the page use clean H1/H2/H3, no decorative headings, lists where appropriate) 7. Freshness signals (does the page show a last-updated date, recent examples, current legislation) 8. Author and expertise signals (is the author named with credentials, linked to a profile) Then give me the top three changes that would move this page the most in AI search results, in priority order. For each, tell me: what to change, why, and roughly how long it would take. When you've done that, ask me if I want to repeat for another page.

Run it on one page. Run it across every service page in turn. Or run it on a competitor's page to see how you compare. For the bigger 18-month "state of the nation" review Mike has commissioned, this is the kind of analysis to ask the website agency to deliver, not bullet-pointed "SEO tips". GEO is what's actually moving traffic now.

The Microsoft Copilot Frontier Program.

Microsoft's early-access channel for Copilot business customers. Tenants enrolled get early visibility on upcoming Copilot capabilities, including Claude models inside Copilot, ahead of general release. Free to apply. Worth Greg's team taking a look if the firm is planning a wider AI rollout, because it materially lifts what agents like the House Style Editor can do without leaving your shutdown Copilot environment.

Microsoft Copilot Frontier Program →