Watch Max build a technographic audience from one prompt.
A real session: companies running Salesforce + Marketo, matched from the live graph — with the team reporting exactly what it could and could not build.
AI agency
Seven AI agents that do the work.
This is not a chat wrapper around dashboards. Metadata agents create audiences, copy, creative, offers, landing pages, budgets, campaign drafts, reporting views, and optimization recommendations for human review. Across 141 MCP tools, the agents can take 140 different actions in your ad channels, CRM, and data warehouse.
END-TO-END
Full overview of what the agents can do.
The team runs the entire paid-media lifecycle, end to end. Flip any stage to meet a customer who proved it.
Strategy
Turn a revenue or pipeline goal into channel mix, account strategy, test plan, and budget groups.
Proof
GoodTime grew pipeline 50% while spending 13% less.
Read the story →
Audience building
Assemble firmographic, technographic, intent, retargeting, and suppression audiences from the 500M graph.
Proof
Cacheflow closed its biggest deal from a cold-targeting MetaMatch audience.
Read the story →
Creative & offers
Generate ad copy, image concepts, video prompts, and offers from your brand kit and proof library.
Proof
Firstup generated $6M in revenue at 122X ROI.
Read the story →
Campaign assembly
Build campaign objects with naming, budget groups, exclusions, UTMs, and approval states.
Proof
BigID launched 300+ campaign experiments on LinkedIn.
Read the story →
Launch
Push approved campaigns live across every channel in one motion.
Proof
Pendo drove $5M in pipeline with a two-person team.
Read the story →
Optimization
Self-optimize every channel continuously, killing losers and scaling winners toward pipeline.
Proof
Zoom cut CPC 77% and tripled ROI with the Bid Agent.
Read the story →
Reporting
Performance views tied to pipeline, opportunity quality, and CAC, not vanity metrics.
Proof
LaunchDarkly fixed attribution and improved cost allocation.
Read the story →
Operations & MCP
Keep integrations, account setup, and governance organized; hand off headless over MCP.
Proof
Docebo aligned sales and marketing with Spotlight.
Read the story →
What the agents produce in the actual product.
The app flow shows audience creation, creative generation, campaign drafting, and feedback while the agent team works.
Prompt-first campaign brief
Start in the LLM-first interface and ask the agent team for the outcome, audience, channel mix, or campaign you need.
Creative generation
Generate ad copy, concepts, visual variants, and offer direction from the brand kit and customer proof library.
Campaign review surface
Inspect campaign objects, launch state, budgets, channels, and approvals in the SaaS interface before spend moves.
Prompt types, tool access, and outputs.
Each agent has a role, but the important buyer question is what it can actually do. These prompts map to public MCP tool families and reviewable outputs.
Mei
Head of StrategyPlans campaigns, coordinates the team, translates executive outcomes into work orders, and keeps approvals moving.
Prompt they handle
Create an enterprise pipeline plan for cybersecurity accounts researching ABM platforms, with channels, audience strategy, offers, and approvals.
MCP tools
get_account_summary_insightsget_insights_reportcreate_audience_from_segmentcreate_campaign
Reviewable outputs
Campaign briefapproval checklistagent task plan
Zoe
Creative DirectorTurns brand DNA, customer proof, offers, and landing-page context into ads for every channel.
Prompt they handle
Generate LinkedIn, Meta, Google, and video creative concepts for a VP Marketing audience using the Zoom and Pendo proof library.
MCP tools
generate_brand_creativegenerate_brand_creative_canvascreate_update_adssearch_library_images_by_name
Reviewable outputs
Ad copyimage conceptsvideo promptslanding-page direction
Max
Audience LeadBuilds firmographic, technographic, intent, retargeting, and suppression audiences from the 500M profile graph.
Prompt they handle
Build an audience of enterprise B2B SaaS companies using HubSpot or Salesforce, researching Demandbase or 6sense, and exclude customers.
MCP tools
create_firmographic_audiencecreate_technographic_audiencecreate_bombora_audienceupload_account_list_csv_audience
Reviewable outputs
Matched audiencesexclusionsretargeting poolsreach estimates
Eva
Offers & Promotions LeadDesigns the offers and promotions behind each campaign: lead magnets, trials, assessments, incentives, and landing-page hooks matched to audience and funnel stage.
Prompt they handle
Propose offers for a mid-funnel enterprise audience — a benchmark report, an ROI assessment, and a trial — with landing-page angles and per-channel variants.
MCP tools
create_offercreate_update_offerfind_offer_urlremove_leadgen_offer_from_campaign
Reviewable outputs
Offer setlanding-page angleschannel variants
Ann
Campaign ManagerAssembles campaigns, attaches audiences and assets, sets budgets, and prepares launch plans for review.
Prompt they handle
Draft a multi-channel campaign with two audiences, three ad concepts, budget groups, UTMs, exclusions, and launch checklist.
MCP tools
create_campaignadd_and_edit_campaign_elementsadd_exclusion_audiences_to_campaignlaunch_campaign
Reviewable outputs
Draft campaignbudget groupsUTMslaunch readiness check
Tom
Analytics DirectorReads performance, identifies winners, and recommends optimizations tied to pipeline and CAC.
Prompt they handle
Find the strongest audience, creative, offer, and channel combinations by pipeline, opportunity quality, and CAC trend.
MCP tools
performance_metricsdeep_funnel_statsexperiment_performance_statsbudget_group_performance
Reviewable outputs
Optimization memowinner/loser listbudget recommendations
Raj
Operations LeadManages integrations, budget groups, account setup, guidelines, and MCP handoff for headless execution.
Prompt they handle
Confirm CRM, MAP, ad channels, brand assets, budget limits, approval rules, and MCP access before launch.
MCP tools
search_campaigns_by_namesget_campaign_by_wizard_idmanage_campaignarchive_audience
Reviewable outputs
Integration statusgovernance checklistMCP handoff notes
What the agents own.
Decision makers get strategic leverage. Operators get fewer repetitive tasks. Every action remains reviewable.
Strategy to task plan
Turn a business goal into channels, segments, offers, tests, budget groups, and launch dependencies.
Audience creation
Find the accounts and people most likely to care using account data, contact data, intent, technology, and CRM context.
Creative production
Produce copy, image ideas, video prompts, cards, banners, and landing-page recommendations consistent with your brand.
Campaign assembly
Build campaign objects with naming, budgets, exclusions, UTMs, approval states, and launch readiness checks.
Optimization cadence
Identify winners and losers faster than a manual team can check every channel.
Operational control
Keep credentials, integrations, account setup, and budget guardrails organized for review and MCP handoff.
One brain, every channel.
Everything your go-to-market generates flows in. The agents and your strategist operate from one seat. Approved work flows back out to every channel, and performance flows back in. The optimization loop never stops.
Everything flows in
Your data, plus a decade of learned spend.
The agents reason over your CRM pipeline and revenue, your marketing-automation leads and journeys, and Metadata proprietary data across 15+ sources with $1B+ in managed spend learned, alongside frontier LLMs and live market signals.
CRM pipeline & revenue
Marketing automation leads & journeys
15+ proprietary data sources
$1B+ managed spend learned
LLMs & market signals
Everything flows back out
One seat, every channel, full control.
From one interface, approved campaigns flow out to LinkedIn, Meta, Google Ads, TikTok, Reddit, X & Bing, programmatic, and ChatGPT Ads. Performance flows back in and the agents optimize against it continuously, with every action reviewable by a human before launch.
LinkedIn
Meta
Google Ads
TikTok
Reddit
X & Bing
Programmatic
ChatGPT Ads
ALWAYS ON
What the agents replace.
No human team or agency can keep up with hundreds of thousands of optimizations a month, and when people leave, their knowledge walks out the door. The agents run 24/7 and remember everything.
Capacity
Teams can't keep up
Manual execution caps out long before the optimizations do. The agents have run 263K+ experiments and self-optimize every channel without burning out or losing the plot.
Memory
Knowledge that stays
The agents kill the experiments that underperform and double down on winners, learning from $1B+ in managed spend. That learning compounds and never leaves with a departing hire.
263K+ experiments$1B+ managed7 patents200+ customersAlways reviewable
Why the old way fails
Your GTM stack was built for a world that no longer exists.
Legacy ABM platforms tell you who to target, then hand the work back to your team. The agents close that gap — live intelligence and autonomous execution in one place.
Legacy ABM platforms
- Lagging intent signals from stale third-party data
- Human-driven dashboards across 5+ disconnected tools
- Manual campaign setup taking days per channel
- Bid and creative optimization by hand
- Intelligence-only: tells you who to target, never runs the campaign
Metadata
- Live buyer intelligence unified from CRM, web, and behavioral signals
- One interface, one prompt, eight channels
- Campaign launch in minutes, not days
- AI agents autonomously optimize bids, budgets, and creative
- Full execution engine: audiences, campaigns, experiments, attribution
Metadata vs. legacy ABM.
| Legacy ABM (6sense, Demandbase) | Metadata | |
|---|---|---|
| Signals | Lagging intent data from third parties | Live, unified buyer intelligence from your own data |
| Execution | Human-driven dashboards | AI-driven, prompt-based campaign execution |
| Optimization | Manual campaign tweaks | Autonomous AI agents optimizing in real time |
| Channels | Display ads only (DSP) | 8 channels: LinkedIn, Meta, Google, Bing, X, Reddit, Instagram, LLM |
| AI / MCP | No MCP server, no agent execution | 141 MCP tools, 7 autonomous agents, any LLM |
| Time to launch | Days to weeks per campaign | Minutes from prompt to live campaign |
“What I really like about the Metadata MCP is that I can now interact and engage and actually get summarized learning of what changes were made and why and how I can learn more about my audience as a result.”
Amanda ColeCMO, Bloomreach
Frequently asked questions
What are Metadata's AI agents?
Seven AI specialists - Mei, Zoe, Max, Eva, Ann, Tom, and Raj - that build audiences, creative, offers, campaigns, budgets, and reporting, then optimize live while your team approves every object.
How many tools can the agents use?
Across 141 MCP tools, the agents take real actions across your ad channels, CRM, and data warehouse.
Do the agents act on their own?
The agents prepare and optimize the work, but your team reviews and approves audiences, creative, budgets, and launches before anything goes live.
Which channels do the agents run?
LinkedIn, Google, Meta, X, Microsoft Advertising, Reddit, and ChatGPT Ads.