An operator's playbook for agencies running competitive intelligence across client portfolios: margin math, per-client qualifying prompts, a 45-minute onboarding template, deliverable formats and pricing tiers.
May 16, 2026 · 14 min read
Agencies that run competitive intelligence for clients have a different problem than in-house teams. The work is technically the same, but the economics destroy any naive setup. Buying a Klue seat per client at $1,500 a month is a margin disaster the moment you cross five clients. Running everything in one shared workspace breaks the multi-tenant promise. Building it manually from scratch eats the practice lead's week.
This guide is for the operator on the agency side: the founder of a digital or GTM consulting agency, the head of a CI practice, or the account director who's been told "add CI to the retainer." It covers the economics, the per-client setup discipline, the deliverable formats that justify the retainer line, and the tools that fit at this scale. For the discipline-level overview, read the competitive intelligence pillar. For the upstream PMM view that some of your clients will hire after they outgrow your service, see competitive intelligence for product marketing managers.
TL;DR
Enterprise CI suites (Klue, Crayon, Kompyte) destroy agency margin above three clients. AI-native workspaces with per-client prompts cost an order of magnitude less and were designed for this exact use case.
Each client needs their own qualifying prompt. One agency-wide prompt produces noise for every client.
A 45-minute onboarding template covers what most agencies do over three meetings: priority competitors, qualifying thesis, deliverable cadence, surface choice.
Three deliverable formats cover most retainers: weekly digest, monthly brief, real-time alerts on high-impact moves.
Price CI as a retainer line item ($500-$3,000/mo) based on competitor count and deliverable cadence, not as a flat fee.
The agency CI economics problem
The economics of running CI at agency scale don't work with enterprise tooling. The math is simple and unforgiving.
Take a sales-led suite at $1,500 per seat per month, which is roughly mid-market pricing for Klue or Crayon. You need one seat per client for proper data isolation. At three clients, your tooling cost is $4,500 a month. To bill that profitably at a 40% gross margin, you need to charge each client $1,875 a month for CI alone, before adding any analyst time. Most agencies bill CI as a retainer add-on at $500 to $1,500 a month. The math doesn't work past three clients.
Two ways agencies usually try to fix this. They run all clients in one shared workspace, which breaks multi-tenant separation and creates data leak risk between competitive clients. Or they cut tooling cost by going to free Visualping plus manual review, which works for two clients and stops scaling above five.
The actual fix is structural. AI-native CI workspaces price per workspace, not per seat, and support multi-tenant isolation by design. Pricing falls to $50-$200 per client per month at this layer. The same margin math at 40% lets you charge $125 to $500 per client and turn a profit. That's the price point clients actually want to pay for a CI line item. It's also the only architecture that scales past ten clients without breaking either the margin or the client-isolation promise.
Ready to monitor your competitors without the manual work?
Competitive intelligence practitioners building AI-native workflows.
This is why the rest of this article assumes AI-native tooling rather than sales-led suites. The role-specific point isn't that enterprise tools are bad; it's that the economics force agencies into a different category by default.
What multi-tenant CI actually requires
"Multi-tenant" gets thrown around loosely. For agencies, four properties matter, and the absence of any one of them creates real problems.
Per-client workspace isolation. Client A's qualifying prompt, watchlist, and signals never appear in client B's surfaces. This is the baseline. Tools that store all CI in one shared database with optional filters fail here because filters get misconfigured eventually.
Per-client qualifying prompt. The thesis defining what counts as a relevant signal differs per client. A SaaS client tracks pricing-page changes; a retailer tracks shelf positioning and ad spend; a B2B service tracks hiring patterns. One prompt across all clients produces noise for every client.
White-label surface. Reports, emails, and dashboards delivered under the agency's brand, not the tool's. Some tools support a logo upload; the better ones support full template control. The difference matters when reports go to client executives.
Per-client access controls. Some clients want to log into the surface directly; others only see the agency's curated output. Multi-tenant tools handle both modes without code; single-tenant workarounds usually break one of them.
When evaluating a tool for agency use, those four are the checklist. Pricing, integrations, signal quality matter, but if multi-tenant isolation is shaky, scaling past three clients is a margin and risk disaster simultaneously.
The per-client qualifying prompt (one brief per client)
The single biggest lever for agency CI work is the per-client qualifying prompt. It is also where most agencies under-invest.
A qualifying prompt is one or two paragraphs answering: what kinds of competitor moves would actually change a decision on this client's side? Generic prompts ("track all major competitor moves") produce dashboards no one reads. Specific prompts ("any pricing tier change in the $50-$200/mo SMB band; any new feature in the integrations vertical; any LinkedIn campaign by their CMO targeting our ICP keywords") produce signals the client acts on within hours.
Three components of a good per-client prompt.
Their decision context. What does the client lose sleep over? Defending a flagship customer segment, fending off a price-war competitor, validating an enterprise positioning shift. The prompt mirrors that lens.
Their stage. A pre-PMM SaaS startup needs different signals than a 200-person company. The prompt names the stage-appropriate signal categories explicitly so generic noise gets filtered out.
Their named competitors. Two to four named priority competitors plus a "watch but don't alert" list of three to five secondary players. The prompt sorts signals by this hierarchy.
Writing the per-client prompt is a 30-minute investment that pays back across every weekly digest. Skipping it produces digests that read the same for every client and signal that the client could have gotten themselves from Google Alerts.
An onboarding template that scales (45 minutes per new client)
The recurring cost of running CI at agency scale is per-client setup. The first time you onboard a client, it takes a day. By the tenth client, it has to take under an hour or the practice doesn't scale.
A repeatable 45-minute onboarding template covers the four decisions that need a human in the loop.
2. Setup (15 min)Workspace, qualifying prompt, sources
3. First delivery (10 min)Tag deliverable cadence, format, contact
4. Cadence lock (5 min)Recurring slot in calendar, monthly review
The brief step is the one to invest in. Fifteen minutes with the client lead to confirm: who are your priority competitors right now, what changed in the last quarter that surprised you, what would you have wanted to know about earlier. Those three questions surface the qualifying thesis better than any structured form.
Setup is execution. The same playbook on every new client: workspace created, prompt written from the brief notes, three to five priority sources configured. If your tool requires more than 15 minutes here, your tool is wrong for agency work.
First delivery confirms the loop is working before the client expects it. A small initial digest within 24 hours validates the qualifying prompt and the deliverable format with the client. Adjust before the first real cadence cycle. The cost of a misaligned prompt rises every week it stays in place.
Cadence lock is the operational discipline. The recurring weekly slot in the client's calendar, the monthly check-in with the account lead, the quarterly review. None of these surprise anyone if locked at onboarding.
Done well, this template makes the per-client setup cost predictable: 45 minutes of senior time, $50-$200 of tooling at the AI-native price point, then ongoing analyst time at whatever cadence the retainer covers.
Deliverable formats that justify the retainer line
Three formats cover most agency CI retainers. Mixing them across clients lets you price by complexity rather than by hour.
Format
Cadence
Audience
Length
Retainer price tier
Real-time alert
Per high-impact signal
Client CMO / CEO / VP product
1-2 sentences in email or Slack
Included or +$200/mo
Weekly digest
Friday afternoon
Client marketing / product team
200-400 words, dated, sourced
$500-$1,000/mo
Monthly brief
First Monday
Client exec, board, all-hands
1-2 page PDF or Notion
+$500-$1,500/mo
Quarterly deep-dive
End of quarter
Client board or strategy committee
8-15 page report
+$1,500-$3,000 one-off
The real-time alert is the value retention layer. Even when nothing in the weekly digest catches the client's eye, one well-timed alert per quarter (a major competitor pricing change, a fundraising event) reminds them why they pay for the service. Skip this and clients churn.
The weekly digest is the bread-and-butter deliverable. It needs to be readable in three minutes, structured the same way every week (consistent format compounds), and dated. Inconsistent digests are the leading cause of agency CI retainer cancellations.
The monthly brief is what the client's leadership sees. It needs to be more analytical than the weekly digests; it should answer "what does this quarter look like competitively" not "what did Crayon's pricing page do this week." This is where senior agency time goes.
The quarterly deep-dive is the highest-priced and least frequent. It's also the format that converts retainer clients into project clients for adjacent work (positioning audits, GTM consulting).
Tools that fit agency economics
For an agency at scale, the tool decision is dominated by per-client cost and multi-tenant capability. Other features matter less than they look in vendor demos.
Layer
Examples
Per-client cost
Fits agencies because
AI-native CI workspace
watchr, newer entrants
$50-$200/mo
Per-workspace pricing, native multi-tenant, MCP/API for white-label routing
Page-change detection
Visualping, Wachete
$10-$30/mo
Cheap baseline signal source per client
AI assistant
Claude, ChatGPT enterprise
$30-$50/mo shared
Synthesis and brief drafting, amortized across all clients
Sales-led CI suite
Klue, Crayon, Kompyte
$1,000+/mo
Only fits when one client demands enterprise tooling and pays for it
The default agency stack in 2026 is the first three. An AI-native CI workspace per client, a free or cheap page-change watcher as a baseline, and the agency's existing AI assistant subscription for synthesis. Total monthly cost per client lands between $100 and $300, which is the price band where retainer CI works.
The sales-led suite only shows up when a specific client demands it. Some enterprise clients have an existing Klue or Crayon contract and want the agency's CI to integrate with it. That's a per-client exception, not the default tooling decision.
Notable absence from the stack: dashboards. Custom client dashboards are tempting to build because tools make it cheap, but clients rarely use them. The deliverable formats above (alert, digest, brief, deep-dive) cover most of what dashboards would deliver, with less ongoing maintenance.
Pricing CI as a retainer line item
Agencies that price CI as a flat retainer fee leave money on the table on enterprise clients and lose money on small clients. Three-tier pricing based on competitor count and deliverable cadence works better.
Tier
Competitors tracked
Deliverables
Monthly price
Starter
Up to 3
Weekly digest only
$500-$800
Standard
Up to 6
Weekly digest + monthly brief + real-time alerts
$1,200-$1,800
Premium
Up to 10
All above plus quarterly deep-dive plus monthly account call
$2,500-$3,500
The competitor count is the right scaling axis because it maps to actual signal volume and analyst time. Doubling competitors doesn't double the work (shared sources, shared synthesis) but it does add roughly 50% per tier, which is what the pricing reflects.
Quarterly deep-dives are usually priced out as one-off projects ($1,500-$3,000 each) rather than included in the retainer, because they require disproportionate senior time. Bundle them for Premium tier or sell separately to Starter / Standard clients.
The mistake to avoid: pricing CI as a flat per-hour service. CI scales like SaaS at the tool layer (per-workspace) plus a small fixed analyst overhead. Per-hour pricing forces the agency into a worse economic model than the underlying cost structure supports.
Anti-patterns specific to agencies
Five patterns hurt agency CI practices at scale.
The one-workspace-for-all-clients pattern. Buying one enterprise tool seat and shoving all clients into it. Cheaper in month one, but creates data leak risk and qualifying-prompt drift because one prompt can't fit ten clients.
The custom-everything pattern. Building a custom dashboard and bespoke report template per client. Sounds client-friendly. Kills margin because every retainer renewal renegotiates the deliverables. Standardize the formats and let clients pick from the tier menu.
The senior-analyst-everywhere pattern. Having the practice lead personally write every weekly digest. Doesn't scale past five clients. The weekly digest should be writable by a junior analyst from the tool's qualified output in 30 minutes per client.
The monthly-renegotiation pattern. Letting clients tweak scope, add competitors, change deliverable cadence outside the contracted tier. Each tweak feels small and reasonable. The cumulative effect is that the retainer no longer matches the work, margin disappears, and the team can't predict load. Lock scope at quarterly reviews, not in ad-hoc requests.
The "we'll fix tooling at scale" pattern. Building the practice on free tools and a Notion board, then realizing at 8 clients that you need a real workspace and the migration is a month of work. Pick the AI-native workspace layer at client two, before the migration cost compounds.
Next steps
If you're building a CI service line at an agency, three moves in the first month outperform anything else.
Pick the AI-native CI workspace before signing the second client. Migration cost from a hand-rolled stack at client five is a month of senior time you don't have.
Write the 45-minute onboarding playbook explicitly. Brief template, prompt template, deliverable format menu. Treat it as a productized SOP.
Set the three-tier pricing. Starter / Standard / Premium. Map each to competitor count and deliverable cadence. Resist tier customization for individual clients.
If you want a CI workspace built for multi-tenant agency economics, watchr is free during the open beta. The per-client qualifying prompt, MCP-readable feed for AI-augmented delivery, and per-workspace pricing are designed for exactly the use case this article describes.
FAQ
What's the minimum number of clients before CI is worth productizing as an agency service?
Two. Below two, it's an ad-hoc consulting engagement, not a productized service. From two clients, the multi-tenant tooling pays for itself and the operational discipline (templates, tiers, cadence) starts compounding. Below two, skip productization.
Can we run CI for our own agency alongside CI for clients?
Yes, and you should. The agency's own CI work (competitors in your client-acquisition space, positioning shifts in your category) gives you the operating context to brief client work. Use the same tooling stack with a dedicated agency workspace.
Should we sell CI as a standalone retainer or bundle it into a broader marketing retainer?
Both work. Standalone retainer is cleaner economically and easier to upsell. Bundled into a broader retainer gives stickier client relationships at slightly worse unit margins. Start standalone; bundle later for strategic clients.
How do we deal with clients who compete with each other on our roster?
Multi-tenant tooling solves the data side. The remaining question is account-team conflict: same lead working both? At small agency size this is unavoidable; at scale, designate primary leads per industry vertical and have them not work on competing accounts.
What's the right cadence for the monthly client check-in?
30 minutes once a month. Walk through the prior month's digests, validate the qualifying prompt is still aligned, surface any decisions the client needs CI to support next month. Skip this and clients silently lose context, and renewal becomes harder.
How long until a CI retainer pays back the onboarding investment?
At the AI-native tooling cost point, a $1,000/mo retainer pays back the 45-minute onboarding plus first-month senior time in week one of the second month. From there, gross margin runs at 60-70% with the right tier mix.
Is there a market for white-label CI as a service for non-marketing agencies (PR, M&A, consulting)?
Yes, and underserved. PR agencies want CI for client positioning monitoring; M&A boutiques want competitor reads on diligence targets; strategy consultants want CI as a recurring layer under positioning projects. The same retainer model and tier structure apply with minor adaptation.
