Skip to content

AI Advisor — Account Manager Guide

How to get the most out of Roam Prospect and the Dashboard AI Advisor.


You have one brain with two front doors.

Roam ProspectDashboard AI Advisor
WhereRo.am chat, any channelThe AI panel inside the client Dashboard
Best forProspecting + meeting prep + ad-hoc market / rank questionsDeep questions on an already-open client
Context it loadsOnly what you tell it in chatEverything the Dashboard has already pulled for the open account
Tools60+ specialised tools (full list) — CRM, deals, ads, GBP, GSC, Ahrefs, LRC grid, InSites, SayMore, HubSpot, company knowledge, Irish places, Meta reach, case studies, build pipeline, deal patterns, competitor resolution, per-AM templates, Creator StudioSame tools; plus all the account’s loaded data up-front
MemoryRemembers within one Ro.am chatRemembers within one Dashboard session

Simple rule:

  • Don’t have the account open? → ask Roam.
  • Looking at the account in the Dashboard? → ask the Dashboard Advisor.
  • Prospecting a business you’ve never worked with? → Roam, every time.
DoorWhat it isWhen to reach for it
HubSpot cardEmbedded panel inside the HubSpot deal recordDeal-in-progress questions; voice dictate (Whisper) built in
Snap Chrome extensionDrag a competitor’s ad/SERP image onto the iconAd-evidence prospecting — uploads to R2, fires the advisor, DMs the rep
Creator StudioJWT’d hand-off door for posts + on-brand imagesType studio (or studio <subscriberId>) in Roam; chat-dropped images come with you

All four front-end doors talk to the same brain — same tools, same data, same guardrails.


You’re meeting a business that isn’t an FCR client yet, or you’re researching a trade/area to find prospects. This is where Roam shines.

A1. Quick rank lookup — the daily driver

Section titled “A1. Quick rank lookup — the daily driver”

These are one-line questions that return in ~10 seconds. These are the real wins — keep them in your head as muscle memory.

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

How does Fergal Colbert Tree Care rank for tree surgeon in Youghal?

Roam will:

  1. Find the business on Google Maps (confirms name + place_id)
  2. Run a 25-point SERP grid over Youghal for “tree surgeon”
  3. Reply with a rank summary: “Rank X/25 points, best Y, avg Z. In top-3 at N points.”
  4. Surface the top competitors that appear across the grid

If the business is based outside Youghal you’ll get a cross-location read — can be useful to see catchment strength.

Identify the primary local competitors to [business] in [town].

Identify three primary local competitors to Fergal Colbert Tree Care in Dungarvan.

Roam will run the SERP grid, aggregate who tops each point across the grid, and return the 3 most-present names. Follow up with “enrich the top three” and it calls enrich_competitors for services, photo counts, response rate, opening hours, and — critically — whether any of them are already FCR clients (so you don’t pitch against your own).

If you’re preparing to meet a trade or category cold, lead with:

Meeting a plumber in Cork tomorrow — what should I know?
Meeting prep: electrician in Louth
Prospect intel for pet shops in Cork

Roam calls prospect_intel and comes back with six sections:

  1. Similar FCR-built sites (sector + location; gold-standard NPS clients first with their review quote)
  2. Search terms — split into:
    • adsSearchTerms — real Google Ads clicks + CPC, lead with these for ads pitches
    • highVolumeTerms — aggregate demand (note volume but don’t pitch against informational “rip.ie”-style queries)
  3. National keywords — Ahrefs monthly volumes, Ireland-wide
  4. Local keywords — Google Keyword Planner, county-level
  5. Area intelligence — population, households, purchasing power, active companies
  6. Consumer review priorities — what customers actually care about in that trade
Cork City
Dungarvan
area map for Naas

Get a full-screen interactive map: small-area demographics, commercial properties, housing, FB/IG ad-reachable audience, new builds. Great as a visual to open during a meeting or paste into a proposal.

A4. Local Rank Check (LRC) — the grid visual the client sees

Section titled “A4. Local Rank Check (LRC) — the grid visual the client sees”
Run an LRC for Fergal Colbert Tree Care in Dungarvan
Run a multi-grid LRC for Fergal Colbert — home, Kilmacthomas, Youghal

Roam will:

  1. Call suggest_insites_params and propose 5 keywords + 1-3 locations (pulling from your KI if they’re a client, their homepage + Irish-places gazetteer if they’re a prospect)
  2. Show you the suggestion list and wait for your confirmation — correct the keywords/locations, add/remove, then say “go”
  3. Submit to InSites via SayMore
  4. Poll for results and come back with the grid + share URL

Never skip the confirm step. Every fabricated keyword burns 25 SerpAPI points and produces garbage for the AM.

Once a grid is run, ask:

Publish the map
Share the map link

Roam publishes a 30-day interactive grid map URL with demographics overlays, a “Competitor” tab, and an InSites iframe drill-down. Prospect-safe — the FCR-client badge is stripped server-side so it won’t accidentally tell a prospect “Acme down the road is our client.”

A6. Case studies — proof points for pitches

Section titled “A6. Case studies — proof points for pitches”
Case studies for an accountant booking system
Do we have a similar success for an e-commerce pet shop?

Roam pulls from the 15 curated case studies authored by the Commercial Director. Each comes back with the business name, account manager, MRR, and the narrative quoted verbatim — don’t paraphrase the Commercial Director’s voice in outreach.

If you ask “list all the case studies,” Roam will tell you how many it returned vs how many exist — reconcile before presenting a partial list.

A7. InSites Digital Footprint (website audit)

Section titled “A7. InSites Digital Footprint (website audit)”
Run a DF on forkliftandpallettrucks.ie
Submit an InSites audit for this URL

DF is the one-page website audit (SSL, mobile, organic, GBP, reviews, GDPR, AI readiness, content, social, tech, indexability, paid search, ecommerce, booking, live chat, backlinks). Roam waits for you to confirm the suggested target keywords, submits, and summarises when the report is back.


Part B — Existing Client Use (in Dashboard or Roam)

Section titled “Part B — Existing Client Use (in Dashboard or Roam)”

You have a subscriber ID (or you’ve clicked on the account). The Advisor already has the full enrichment loaded: CRM, products, MRR, GBP, keywords, ads, GSC, Ahrefs, tickets, call tracking, reviews.

B1. Quick lookups — no opening, just ask

Section titled “B1. Quick lookups — no opening, just ask”
When's the next bill for Valentine Ladders?
Any open tickets on 38007492?
What products does Brian Stephens' portfolio include?
Is Maura Mackey Design suspended?
How much MRR does this account have?

These are instant — the /check-account data is already on screen.

B2. Performance deep dive — the “how are we doing” question

Section titled “B2. Performance deep dive — the “how are we doing” question”
How's Google Ads performing this month?
SEO detail — what tracked keywords moved?
Any performance drops I should flag to the client?

For ads, the Advisor will always:

  • Quote pre-calculated CPA / CPC / CTR — never calculate itself
  • Break Search campaigns out individually (not rolled into one SEARCH bucket)
  • Flag Call-Only campaigns separately
  • Apply the PMax caveat: PMax cost-per-conversion is not directly comparable to Search — includes Shopping, GBP actions, map clicks, view-throughs. Never recommend budget shift without a conversion-action parity check + client lead-quality check.
  • Flag the conversion mix — if all conversions are click-to-call and no form/purchase tracking, that’s a measurement gap worth fixing.

For SEO clients (anyone with SEO Plus / Optimiser / MapRank in their program), the Advisor leads with:

  • Ahrefs tracked keywords (what FCR committed to moving) with position, change, landing page
  • Backlink diagnostics (DR, dofollow ratio, spam-farm flags)
  • GSC impression-trap (keywords showing but below click threshold)
  • Not the KI gap list — that’s for prospects; for paying clients the tracked set is the accountability surface
Deep dive on Pro Fix Roofing
Full review of 38007258 — I've got a QBR Thursday

Returns:

  • Active program + MRR + renewal dates
  • DEA / Setup service periods (fromDatetoDate, 30/60/90-day flags)
  • Google Ads, GA4, GSC, GBP, call tracking (if installed), SEO, reviews
  • Keyword capture vs category
  • Risks (suspended lines, aged tickets, declining trends, low reviews)
  • Opportunities (missing products, missing GBP services, uncaptured keywords)
  • Health score breakdown

Use this the day before a QBR.

What's missing from Fergal Colbert's program?
Upsell angles for this account

The Advisor reads the active product list and surfaces:

  • No website built by FCR → SitePro / LocalRank
  • No listings management → SayMore Ultimate
  • Retailer with no shopping → SayMore Retail
  • Phone-heavy business with no call tracking → iovox
  • Website with low organic → SEO
  • No Google Ads in a high-volume category → SEA

It won’t pitch a product the client already has — and it won’t say a prospect “doesn’t have X” based only on our own data (FCR covers ~3-5% of the market).

B5. Follow-up questions (conversation memory)

Section titled “B5. Follow-up questions (conversation memory)”

Within a single chat, Roam and the Dashboard Advisor both remember context. After a deep dive, you can keep digging:

[after a deep dive]
What about their GA4 traffic mix?
Run an LRC on that
Tell me more about that Search campaign
Who are their top 5 competitors?
Publish the meeting brief

If you switch topic (a different client / a different category), the Advisor will start fresh — don’t try to carry history across unrelated accounts.

B6. Pick up a brief / map a colleague shared with you (context bridging)

Section titled “B6. Pick up a brief / map a colleague shared with you (context bridging)”

When someone forwards you an advisor output — a meeting brief or grid map (/proposal/… link) or a discovery report (/r/… link) — you don’t have to re-enter anything to keep working on it. Paste the link into a fresh chat and the Advisor re-loads its saved context (business, category, location, subscriber / deal, keywords, grid labels, plus who made it and when), confirms what it loaded, and lets you continue:

[paste the /proposal/ or /r/ link]
→ "This is the meeting brief for Laois Jetwash (Drain Cleaning, Portlaoise) that Cathal ran on 9 Jun. Continue from here?"
Run an LRC on this
Draft a proposal
Save this as a meeting brief under my name

This works across users (the link Cathal created, you continue) with no session token. Briefs and grid maps are kept for 30 days — if a shared link has expired, the Advisor says so and offers to rebuild it from the business name or deal. Discovery reports (/r/…) don’t expire.


Gold Patterns — prompts that consistently deliver

Section titled “Gold Patterns — prompts that consistently deliver”

Keep these in your back pocket. They reliably trigger the best tool chain.

GoalPrompt
Rank spot-checkHow does [business] rank for [keyword] in [town]?
Competitor ID (fast)Who are [business]'s competitors? — Ahrefs organic rivals + local map-pack + FCR-client flags in one call (~15s)
Competitor ID (ground-truth rank)Identify three primary local competitors to [business] in [town] or Run an LRC for [business] in [town] — 25-point grid + shareable map artefact
Meeting prep (cold)Meeting a [trade] in [town] tomorrow — what should I know?
Market read (trade-level)Prospect intel for [trade] in [county]
Area snapshot[Town name] (just the name — triggers area map)
LRC submissionRun an LRC for [business] in [town]
Grid publishPublish the map (after a grid has been run)
Deep diveDeep dive on [business] / [subscriber ID]
SEO detail (paying client)SEO detail — what tracked keywords moved?
Ads detailHow are the ads performing this month?
Account basicsWhen's the next bill / Any open tickets / What products for [X]
Similar-client proofDo we have a case study for [sector / solution]?
Website auditRun a DF on [url]

  1. Calculate metrics itself. CPA, CTR, conversion rate, averages — it quotes pre-calculated values or says “not available.” Never divides spend by conversions in its head. If a number looks off, it’s the data, not the model.
  2. Say a prospect “has no Google Ads” / “no SEO.” Our data covers ~3-5% of the Irish SME market — absence in our system doesn’t mean absence in reality.
  3. Compare PMax vs Search conversions as equals. It’ll always attach the parity + lead-quality caveats before any budget-shift recommendation.
  4. Submit an InSites audit without your confirmation. It proposes, you sign off, it submits.
  5. Invent case studies. If a case study isn’t in the curated set, it’ll say so.
  6. Present a 5-of-15 case-study search as “the complete list.” It reconciles against the index total.
  7. Publish a grid map for a different business than the one the grid was run for. The cache is target-scoped; mismatch = refuse.

  • Vague ask = thin answer. “Check this ID and see where we can improve SEO” will get you a generic dive. “Pull the Ahrefs tracked keywords for Pro Fix Roofing and flag what’s moved in the last 60 days” gets you the focused answer the first time.
  • Reusing a grid across prospects in one session. If you ran a grid for Prospect A and then switch to Prospect B, re-run the grid. Ask for “publish the map” without re-running and the Advisor will now refuse — but muscle memory is: new business, new grid.
  • Acting on conversion numbers without reading the caveat. If the Advisor flags “only click-to-call tracked,” the “conversions” line in the ads data is phone taps, not leads. Treat it accordingly.
  • Asking Roam a question about an account that’s open in the Dashboard. The Dashboard already has the enrichment loaded — asking Roam forces it to re-fetch. Faster and richer to stay in the Dashboard panel.
  • Copy-pasting case study narratives and “fixing” typos. The Commercial Director’s voice is the feature. Quote verbatim.
  • Treating the PMax CPA as comparable to Search CPA. Don’t.

  1. Lead with the subscriber ID when you have it — the Advisor pins on the ID for all downstream tool calls. No guessing.
  2. State the output shape you want — “give me a three-row table” / “one paragraph summary” / “bullet list of risks.” The Advisor will respect it.
  3. Say what the meeting or decision is“I’m re-pitching at renewal next Thursday” changes the framing. The Advisor will prioritise risks + conversion hooks over background.
  4. Re-ask with a sharper scope if an answer sprawls. Follow-up questions are cheap.
  5. Use “near me” — never a locality name — when asking about searcher intent unless you’re deliberately testing a locality-embedded keyword.
  6. For a second opinion, ask the same question in the other advisor. Roam fetches fresh from tools; Dashboard uses loaded context. If they disagree, the newer data (Roam) is usually right.

  • Something looks wrong → flag the specific prompt + output to Cathal. The system prompt and tools are tuned weekly based on real AM sessions.
  • Tool returned nothing when you expected data → the data may be a genuine gap (small account, no Ahrefs, no GSC). The Advisor will say so rather than invent.
  • Stuck on a prompt shape → start with a Gold Pattern above and follow up.

68 tools, one reasoning engine. You don’t need to memorise them — ask the question, the Advisor picks the combination. Listed here so you know what’s possible, and so the deck has a single source of truth to track against.

1. Account intelligence — one client, who/what/when (15)

Section titled “1. Account intelligence — one client, who/what/when (15)”

check_account · get_enrichment · get_portfolio · get_product_history · get_deals · get_deal_by_id · get_deal_brief · find_similar_deals · extract_deal_patterns · get_tickets · get_ticket_by_id · post_note_to_ticket · get_projects · get_project_by_id · get_suspensions

2. Performance & channel data — how the marketing is doing (14)

Section titled “2. Performance & channel data — how the marketing is doing (14)”
  • Ads: get_ads_stats · get_ads_quality · get_ads_portfolio · get_ads_accounts_no_conversions
  • Search / SEO: get_gsc · get_ahrefs · get_ahrefs_by_url · get_keywords · get_sitepro
  • Local presence: get_gbp · get_gbp_live · get_gmc · get_reviews
  • Calls: get_call_tracking

3. Rank, competitor discovery & audits (11)

Section titled “3. Rank, competitor discovery & audits (11)”

search_google_places · run_serp_grid · check_satellite_overlap · enrich_competitors · publish_grid_map · submit_insites_df · submit_insites_lrc · get_insites_report · get_insites_audit_settings · suggest_insites_params · generate_insites_share_link

  • prospect_intel — similar FCR-built sites + commercial search terms + Ahrefs/Keyword Planner volumes + area demographics + consumer review priorities. New 2026-05-26: the competitors:true mode adds an Ahrefs organic-competitors block (share-filtered, aggregator-denylisted) + a Google Places map-pack (FCR-client-flagged) — the canonical fast path for “who are X’s competitors?”
  • crawl_website — home-page features (booking / ecommerce / blog / gallery), tech stack, payments, AI-readiness, sitemap depth

get_revenue_bridge · get_am_bridge · get_sales_rep_bridge · get_am_attributes · get_build_summary · get_build_pipeline · get_build_detail

search_company_knowledge — 15 curated case studies (Commercial Director’s voice, quoted verbatim), positioning copy, marketing-curated FAQs / proof points / value props. Standalone marketing curation skills (/marketing-knowledge, /marketing-products, /marketing-outreach) keep this index live.

7. Outreach, sales actions & write-backs (10)

Section titled “7. Outreach, sales actions & write-backs (10)”
  • Template library: find_outreach_template · get_am_email_templates · list_team_email_templates · save_my_email_template · save_team_email_template · delete_my_email_template · delete_team_email_template
  • Deal write-backs: save_proposal_to_deal · log_call_to_deal · send_brief_to_development

save_my_preference · list_my_preferences · forget_preference — stored in KV with no TTL, surfaced as a standing-preferences addendum to every turn. The Advisor will ask before saving and will list / forget on request.

  • generate_image — Ideogram via FCR Creator (chat-time one-off images)
  • open_studio ★ — mints a Studio JWT and carries your chat-dropped images straight into Creator Studio (logo + base composite, posts, on-brand graphics). Type studio or studio <subscriberId> in Roam.
  • run_bigquery — ad-hoc SQL across SUPERSET_MASTER / fcr_operations. 30 GB billed-bytes runtime cap + Ro.am DM alarm on overrun.
  • read_google_sheet — read any Google Sheet shared with fcr-dashboard@listingmanager-1529856313699.iam.gserviceaccount.com (Viewer). Values only, no formatting.
  • save_meeting_brief — publish a shareable Cloudflare-hosted brief (markdown → HTML, 30-day TTL).

Capabilities you can ask for that aren’t a single tool

Section titled “Capabilities you can ask for that aren’t a single tool”
  • Voice dictate in the HubSpot card — Whisper-backed /voice-stt.
  • Studio door — bare studio (with optional subscriber ID) returns a JWT’d link straight into Creator Studio, skipping the LLM.
  • Area-map summary — type just a town name (e.g. Youghal, Donnybrook) and the Advisor returns small-area demographics + reachable audience + companies, with the live area-map URL inline.
  • The grid-map artefactpublish the map after a run_serp_grid produces a prospect-safe Cloudflare URL: rank pins, demographics tabs, Web Results lightbox, draw-a-circle, InSites iframe drill-down. FCR-client flags stripped server-side.
  • Context bridging — paste a shared brief / grid-map (/proposal/…) or report (/r/…) link into a new chat and the Advisor re-hydrates its saved context (cross-user, no session token) so you can keep working on a colleague’s output without re-typing. Expired links (30-day TTL) are flagged with an offer to rebuild. See B6.
  • Per-AM email-template library — personal templates stacked over team-shared ones. Saved templates surface back through find_outreach_template ranked by intent.
  • Guardrails (deterministic, not just prompted): em-dash strip, citation-URL rules for HubSpot / Teamwork / Desk, the “never invent a competitor or external URL” rule with a DoH link-check gate on client-facing drafts.

The deck’s Appendix A listed 27 tools across six categories. The real count today is 68. The biggest deltas:

  • Today (2026-05-26): competitor resolution in prospect_intel (Ahrefs organic + Places map-pack + FCR flags) with a clean handoff to run_serp_grid for local-rank / LRC / shareable-map asks · the open_studio Creator Studio door.
  • Team / cross-portfolio added since the deck: the AM, rep, and revenue corkscrew bridges; the website-build pipeline tools; the suspensions report.
  • Sales actions added: the template library (personal + team), save_proposal_to_deal, log_call_to_deal, send_brief_to_development, post_note_to_ticket.
  • Per-AM durable memory is new — preferences persist across sessions.
  • Power-user data added: read_google_sheet, get_ahrefs_by_url (any domain, not just subscribers), extract_deal_patterns, find_similar_deals, check_satellite_overlap.

This appendix is the source of truth — if the deck disagrees, this wins. Updated as tools are added (regenerated from the tools[] array in worker/src/handlers/ai-advisor.js).

Ask the docsRAG over this site
Ask anything about the FCR Dashboard platform — architecture, BigQuery, the worker routes, billing rules, the LRC stack, scoring… Answers are grounded in this documentation, with source links.
How does the deal-brief refresh work? Which routes are Worker vs n8n? How is account health scored?