AI Advisor — One Brain, Two Front Doors (v1 deck, April 2026)

FCR Media — Internal. The original presentation on what the AI Advisor does, prepared for Sales, Account Managers, CS, Campaign Managers (v1, April 2026). Transcribed here as the historical origin of the platform — we've since shipped the HubSpot card, precomputed deal briefs, the maps/LRC stack and more (see the playbooks for current functionality). Living source of truth: AI_ADVISOR_GUIDE.md.

The mental model — one AI brain, two front doors

• Roam — in chat, on any device. Tell it what to look at.
• Dashboard — in-account, in the Dashboard. The enrichment is already loaded.
• Shared: 20+ tools, one reasoning engine.

Both answer prospect and existing-client questions — the difference is where you're starting from. (More doors on the way — see the Roadmap.)

Roam vs Dashboard — where you are decides which door

Not whether the business is a client — just where you're starting.

Why it matters

Collapses fifteen-minute lookups into sixty-second answers, and moves AMs from "I'll get back to you" to "I already checked."

• ~60s from a one-line question to a 25-point SERP grid with rank summary + top competitors.
• 20+ tools the Advisor can reach.
• 15 curated FCR case studies it can cite — business, AM, MRR, narrative in the Commercial Director's own voice.

Hero prompts

1 — The sixty-second rank check

"How does [business] rank for [keyword] in [town]?"

Resolve (find on Google Maps, confirm place_id) → Grid (25-point SERP grid) → Summary (rank X/25, best Y, avg Z, top-3 at N) → Competitors (names across the grid). The daily driver — keep this shape as muscle memory.

2 — Who are we up against?

"Identify three primary local competitors to [business] in [town]."

Grid → aggregate who tops each point → top three → follow up with "enrich the top three" (services, photos, FCR-client flags). You walk in knowing the three names the client will mention — and which you already own.

Part A — Prospect use (Roam, cold to close)

• A1 Quick rank lookup — the daily driver, ~60s: grid + rank summary + top competitors.
• A2 Market intel — cold meeting prep: similar FCR-built sites, ads terms + CPC, national & local keywords, area intelligence, consumer review priorities.
• A3 Area map — the visual the client sees: small-area demographics, commercial properties, FB/IG reachable audience, new builds.
• A4 Local Rank Check — suggests keywords + locations, waits for sign-off, submits.

Part B — Existing client use (Dashboard, file open)

• B1 Quick lookups — next bill, open tickets, MRR, product mix, suspended lines (check-account already on screen).
• B2 Performance deep dive — ads/SEO with caveats: pre-calculated CPA/CTR, Search broken out, Call-Only flagged, PMax parity caveat before any budget-shift rec.
• B3 Full client deep dive — the day before a QBR: program + MRR + renewals, DEA periods, ads/GA4/GSC/GBP, call tracking, reviews, risks, opportunities, health score.

Gold patterns (screenshot this)

Guardrails — by design (the refusals are features)

• Numerical discipline: won't calculate CPA/CTR/conversion itself — quotes pre-calculated values or says "not available." Won't treat PMax CPA as comparable to Search without parity + lead-quality checks. Won't recommend a budget shift without the caveats.
• Epistemic discipline: won't say a prospect "has no Google Ads" — our data covers ~3–5% of the Irish SME market. Won't invent a case study.

Common pitfalls

• Vague ask = thin answer — name the timeframe, surface, and decision.
• Reusing a grid across prospects — new business, new grid (cache is target-scoped).
• Acting on conversion numbers without the caveat (click-to-call = phone taps, not leads).
• Asking Roam about an open-in-Dashboard account (forces a re-fetch).
• Paraphrasing the Commercial Director's case-study voice.

Tips for better answers

1. ID first — lead with the subscriber ID when you have it.
2. Shape the output — "three-row table", "one paragraph", "bullet list of risks".
3. State the decision — "re-pitching at renewal next Thursday" changes the framing.
4. Sharpen & re-ask — follow-ups are cheap.
5. "Near me" — use "near me", not a locality name, where it fits.

Roadmap (as of v1)

• Shipped: Roam, Dashboard.
• Exploring: HubSpot embed (post-contract surface — now shipped), Teams/Slack.
• Watching: GoldenPages.ie OAuth (listing-owner sign-in).

Appendix — everything the advisor can reach (27 tools, one engine)

You don't need the tool names — ask the question, the advisor picks the combination. Account intelligence (check_account, get_enrichment, get_portfolio, get_product_history, get_deals, get_deal_by_id, get_tickets, get_projects), performance data (get_ads_stats, get_ads_quality, get_ads_portfolio, get_ga4, get_gsc, get_gbp, get_gmc, get_call_tracking, get_reviews, get_ahrefs, get_keywords, get_sitepro), and the prospecting/maps/knowledge tools.

Full speaker notes (presenter narration)

Transcribed from the deck's speaker notes — the 'why & how' narration behind each slide. Included so the vector index has the full talk track, not just the on-slide bullets.

Slide 1. Welcome. Over the next ten minutes I want to show you the AI Advisor — what it is, why it matters, and how AMs can actually use it in the next meeting. Two products, one brain: Roam for prospecting, and the FCR Dashboard Advisor for existing clients. Keep that framing in your head.

Slide 2. Here's the whole system in one diagram. One AI brain, two front doors. Roam is the chat — anywhere, any device. Dashboard is the panel inside the client Dashboard, pre-loaded with the full enrichment for whatever account you've got open. Same tools under the hood. Both doors can answer prospect and existing-client questions — the difference is where you're starting from.

Slide 3. This is the split you need to internalise. Where you are decides which door, not whether the business is a client. Roam: you're mobile, you don't have the account open, you give it the context in chat. Dashboard: you're at your desk, the account is already open, the enrichment is on screen before you open your mouth. Rule of thumb: account open? Dashboard. Not open? Roam. You can absolutely use Roam on a client — you just have to name them.

Slide 4. Why this matters commercially. AMs are time-poor. A ten-second rank question used to take fifteen minutes of grid-building. The Advisor gives you the shape of the answer — rank, competitors, and fifteen curated case studies in the Commercial Director's own voice — before the meeting even starts. That's the unlock.

Slide 5. First of two hero prompts. 'How does Fergal Colbert Tree Care rank for tree surgeon in Youghal?' One line, ten seconds. Roam finds the business on Maps, runs a 25-point SERP grid, comes back with rank summary and top competitors across the grid. This is the daily driver — muscle memory.

Slide 6. Second hero prompt. 'Identify three primary local competitors to Fergal Colbert Tree Care in Dungarvan.' Roam runs the grid, aggregates the top names, and — critically — you can follow up with 'enrich the top three' and it'll tell you which ones are already FCR clients, so you don't pitch against your own book.

Slide 7. Full tour of Prospect use cases in Roam. Quick rank lookup, market intelligence for cold meetings, area maps, Local Rank Checks with the grid visual the client sees, prospect-safe publishable maps, and case studies quoted verbatim. Start with the fastest-to-value — the rank lookup — and work down.

Slide 8. Existing-client use cases in Dashboard. Quick account basics — bills, tickets, products, MRR. Performance deep dives on ads and SEO with all the right caveats baked in. Full deep dives before a QBR. Cross-sell angles. And crucially, conversation memory within a session so follow-up questions stay cheap.

Slide 9. This is the artefact to screenshot and pin above your monitor. Gold Patterns. The prompt shapes that consistently trigger the best tool chain. Goal on the left, prompt in monospace on the right, channel tag on the far right. If you learn nothing else from this deck, learn these.

Slide 10. Guardrails. Things the Advisor won't do by design. It won't calculate metrics itself — it quotes pre-calculated values. It won't say a prospect 'has no Google Ads' because we only see three to five percent of the market. It won't treat PMax CPA as equivalent to Search CPA. These aren't bugs; they're safety.

Slide 11. Common pitfalls. Vague asks get thin answers. Don't reuse a grid across different prospects in one session. Don't act on conversion numbers without reading the caveat. Don't ask Roam about an account that's already open in the Dashboard. And never paraphrase the Commercial Director's case study text — quote verbatim.

Slide 12. Six tips that compound. Lead with the subscriber ID when you have it. State the output shape. Say what the meeting or decision is. Re-ask with sharper scope if the answer sprawls. Use 'near me' for searcher-intent testing. And when you want a second opinion, ask the same question in the other door — Roam pulls fresh, Dashboard uses loaded context. If they disagree, the newer data is usually right.

Slide 13. Your homework. Three prompts to try this week. Pick a prospect you're meeting, pick an existing client you've got a QBR with, and run a rank check on someone in your patch. Each one takes under a minute. You'll come back to the next meeting with a better question.

Slide 14. Roadmap. The advisor is one brain; the doors are where we meet the AM in the moment. Today: Roam and Dashboard. Exploring: a HubSpot embed that fires when a deal closes, and a Teams or Slack bot to lower the bar to a one-line rank check. Watching: Golden Pages OAuth so listing owners get the advisor inline, and a weekly email digest per portfolio. Principle — and this is the line to quote back at me: we don't build a door until we see AMs reaching for one.

Slide 15. That's the tour. Content lives in the fcr-dashboard repo, updated weekly by Cathal. If something looks wrong, flag it with the specific prompt and output to Cathal — the system prompt is tuned off real AM sessions. Thanks.

Slide 16. Appendix — for the curious. Twenty-seven tools, grouped by what they're for: account intelligence, performance data, rank and audit, prospect intelligence, knowledge and proof, and power tools. You don't need to memorise the names. Ask the question in plain English and the advisor picks the combination. This slide is here as reference; skip it if the list feels engineering-y.

FCR Dashboard documentation · generated from docs/ · keep counts verified, not guessed.