The landscape · Sales and marketing
Sales and marketing: what AI is really doing
Brief me in 2 minutes
AI in Sales and marketing.
AI agents already draft outbound email, resolve a real share of support tickets, and generate a majority of paid ad creative at scale, that part is not hype. What is oversold is autonomy: the receipts describe agents that draft, monitor, research and update records with a human still approving the risky step, and at least one leading vendor got caught inflating the numbers behind its 'fully autonomous' pitch.
The game: we show you 5 things happening. You call each one real now, early or mostly hype. Then the receipts.
Shift 1 of 5
Customer support agents now get paid per resolved ticket, not per seat
Real now
This is deployed and billed now at large regulated enterprises, Fortune 50 customers and hundreds of millions in committed capital chasing live ARR, not a pilot dressed up as a product.
How it works. Sierra and Decagon sell agents that resolve support conversations end to end, refunds, account changes, technical troubleshooting, without a human touching the ticket. Salesforce answered by bundling the same idea into Agentforce Contact Center, wiring voice, CRM data and agents into one workspace its customers already own.
Overstated: Resolution-rate numbers are vendor-reported and vary hugely by task type; one published average hides how much of it is easy FAQ deflection versus a genuinely complex claim or refund.
Shift 2 of 5
AI answer engines are eating the click, and AEO/GEO is the discipline built to chase it
Real now
The click collapse itself is measured and already happening; the tooling built to respond to it, Profound, Otterly, HubSpot AEO, is brand new and unproven at fixing anything.
How it works. When Google shows an AI Overview, people click a normal search result about half as often as when there is no AI summary, and they are more likely to end the search session entirely. Because the citation inside the AI answer now matters more than the ranking below it, monitoring vendors like Profound and Otterly.ai sell dashboards that track whether and how a brand gets cited in ChatGPT, Gemini and Perplexity, and HubSpot has now built the same tracking directly into its own CRM.
Overstated: GEO vendors sell visibility monitoring, not control; no one, including the vendors, can guarantee a model cites a given brand, and Profound raising $180M to build the dashboard says as much about how unresolved the problem still is as it does about demand for the product.
Shift 3 of 5
Generative ad creative is now the default setting, not the experiment
Real now
60 to 70 percent of managed spend on one platform's AI suite is live budget, not a beta test, even though the marketers running it say it is not better yet.
How it works. Meta's Advantage+ suite, which auto-generates creative variants, targeting and budget allocation, is no longer an opt-in product. Agencies report it already runs 60 to 70 percent of their managed Meta ad spend, learning from live auction performance rather than a creative brief, so the 'ad' is really a constantly regenerating set of variants the algorithm keeps testing against itself.
Overstated: Meta says full automation could arrive by the end of 2026, but the same agencies quoted in the same reporting say their own tests still show the AI creative tool performing worse on some tasks, and call that timeline unrealistic.
Shift 4 of 5
AI SDRs promise a fully autonomous rep, the receipts show a human-monitored draft machine
Early
Clay's enrichment layer is genuinely billed and used at scale; the 'autonomous rep that closes deals' pitch behind 11x and Artisan is still trial-stage, with disputed usage numbers on the record.
How it works. Clay, 11x and Artisan each sell some version of an AI worker that finds leads, enriches contact data, writes personalized emails and books meetings without a rep touching the sequence. Clay has become the data and enrichment backbone that many of these agents plug into, rather than a rep-replacement product itself.
Overstated: The pitch is 'AI closes deals'; the receipts show AI drafting and sending, with a human still approving the sequence, or, in at least one competitor's design, still running the actual sales conversation.
Shift 5 of 5
CRM incumbents and AI-native challengers both want to own the agent that acts on your customer data
Early
The agents are shipped and billed, but the receipts describe monitoring, researching and updating records, not agents left unsupervised at the moment a deal or a customer relationship is actually decided.
How it works. Salesforce (Agentforce) and HubSpot (Breeze, Growth Context) are wiring their own agents directly into the CRM data and workflows they already sell. AI-native challengers like Attio and Rox AI pitch a CRM built around agents from day one, arguing an agent bolted onto a decade-old data model cannot reason about an account the way one built for it can.
Overstated: Every vendor here describes agents that prospect, research and update records; none of the receipts we found describe an agent unsupervised at the point of an actual customer commitment.
The scary part
What happens to jobs in Sales and marketing
What AI is already taking:
- Drafting and sending outbound email sequences end to end, prospecting, enrichment and replies TechCrunch ↗
- Running a unified contact center across voice and digital channels off live CRM data Salesforce ↗
- Resolving complex, high-stakes customer cases, insurance claims, refinancing, order returns, without escalation TechCrunch ↗
- Generating and testing the ad creative variants that most of a brand's paid media budget now runs on Marketing Brew ↗
What stays human:
Closing an enterprise deal, deciding what an angry or churned customer is actually owed, and standing behind a number when a regulator or a board asks where it came from all still run through a person. Every vendor receipt we found describes an agent that drafts, monitors, updates or resolves, not one left alone at the moment something cannot be undone, and Monaco built its entire product around keeping a human in the actual sales conversation for exactly that reason.
Get better · 15 minutes
The ability to audit an AI vendor's own numbers, resolution rate, reply rate, churn, contracted ARR, against the raw underlying data instead of trusting the dashboard, the exact skill that exposed 11x's gap between claimed and real revenue.
Try this this week:
Pick one AI tool your team already pays for, a support agent, an AI SDR, an ad platform's auto-creative, pull its self-reported outcome number for last month, then manually recount a sample of 20 tickets, emails or ad variants against the raw records yourself. Budget 15 minutes; if your count and the dashboard disagree by more than a few points, you have found your real number before it reaches a renewal decision.
Your move
The outbound AI-SDR layer is crowded and cooling after 11x's own numbers unraveled under reporting. The open lane is the boring reliability layer nobody has built yet, an independent auditor that checks an AI agent's claimed resolution rate, reply rate or churn against the vendor's raw logs, the way TechCrunch effectively did to 11x by hand.
Watch whether an outcome-based pricing model survives a second-year renewal at the same rate, that is the real test for the whole customer-support category. Red flag to probe in a pitch: a 'contracted ARR' number that counts trial customers who never converted past 90 days, the exact gap reporters found inside 11x's books.
Pull your own CRM or support tool's per-agent numbers this quarter, resolution rate, reply rate, meetings booked, and reconcile a sample of 20 by hand against the raw tickets or threads. If the dashboard and the manual count disagree by more than a few points, that gap is your real number, not the one in the vendor's deck.
Learn to interrogate an AI vendor's headline metric instead of accepting the dashboard: ask for ticket-level or contract-level data behind any published resolution or churn rate before signing a renewal. Start with whatever support or CRM tool your own company already has an active contract with.
What's actually changing
- 1
Customer support agents now get paid per resolved ticket, not per seat
How it works. Sierra and Decagon sell agents that resolve support conversations end to end, refunds, account changes, technical troubleshooting, without a human touching the ticket. Salesforce answered by bundling the same idea into Agentforce Contact Center, wiring voice, CRM data and agents into one workspace its customers already own.
Why it matters. Contact-centre headcount growth slows wherever a resolution rate is trusted enough to act on, and the commercial fight moves to who gets to define and audit 'resolved,' since that definition is what gets billed.
What's overstatedResolution-rate numbers are vendor-reported and vary hugely by task type; one published average hides how much of it is easy FAQ deflection versus a genuinely complex claim or refund.
SierraDecagonSalesforceIntercom
3 receipts
- Sierra raised $950M at a valuation above $15B and says its agents now handle full end-to-end cases, including insurance claims and returns, for over 40% of the Fortune 50.
billions of interactions, from refinancing mortgages to processing insurance claims, managing returns, and powering nonprofit fundraising campaigns
TechCrunch, 4 May 2026 ↗ - Decagon tripled its valuation to $4.5B within six months on demand for its consumer-facing concierge agents.
Decagon has raised a fresh $250 million in funding, led by new investors Coatue Management and Index Ventures, tripling our valuation to $4.5 billion in under six months.
Decagon, 27 Jan 2026 ↗ - Salesforce built Agentforce Contact Center to unify voice, digital channels, CRM data and AI agents in one system.
the only AI contact center solution that unifies voice, digital channels, CRM data, and AI agents natively in a single system
Salesforce, 10 Mar 2026 ↗
- Sierra raised $950M at a valuation above $15B and says its agents now handle full end-to-end cases, including insurance claims and returns, for over 40% of the Fortune 50.
- 2
AI answer engines are eating the click, and AEO/GEO is the discipline built to chase it
How it works. When Google shows an AI Overview, people click a normal search result about half as often as when there is no AI summary, and they are more likely to end the search session entirely. Because the citation inside the AI answer now matters more than the ranking below it, monitoring vendors like Profound and Otterly.ai sell dashboards that track whether and how a brand gets cited in ChatGPT, Gemini and Perplexity, and HubSpot has now built the same tracking directly into its own CRM.
Why it matters. Organic traffic to informational content keeps shrinking even as transactional, purchase-intent searches keep clicking through, so marketing teams are shifting spend and attention toward being the source an AI model quotes rather than the page a person visits.
What's overstatedGEO vendors sell visibility monitoring, not control; no one, including the vendors, can guarantee a model cites a given brand, and Profound raising $180M to build the dashboard says as much about how unresolved the problem still is as it does about demand for the product.
GoogleOpenAIProfoundOtterly.aiHubSpot
3 receipts
- Pew's own browsing-panel study found people click a normal search result about half as often when an AI summary is present.
Users who encountered an AI summary clicked on a traditional search result link in 8% of all visits.
Pew Research Center, 22 July 2025 ↗ - HubSpot shipped its own answer-engine-optimization product built directly into the CRM.
HubSpot AEO is a new solution for answer engine optimization that helps marketers tap into that opportunity by understanding, tracking, and optimizing how their business appears in answer engines like ChatGPT, Gemini and Perplexity.
HubSpot, 14 Apr 2026 ↗ - Profound raised $180M at a $1.8B valuation to build the analytics layer for AI-search visibility.
Profound today announced that it has raised $180 million in funding at a $1.8 billion valuation.
SiliconANGLE, 15 Sept 2026 ↗
- Pew's own browsing-panel study found people click a normal search result about half as often when an AI summary is present.
- 3
Generative ad creative is now the default setting, not the experiment
How it works. Meta's Advantage+ suite, which auto-generates creative variants, targeting and budget allocation, is no longer an opt-in product. Agencies report it already runs 60 to 70 percent of their managed Meta ad spend, learning from live auction performance rather than a creative brief, so the 'ad' is really a constantly regenerating set of variants the algorithm keeps testing against itself.
Why it matters. Ad production shifts from an agency line item into the ad platform's own black box, good for speed, bad for anyone whose job was making the creative decisions the algorithm now makes instead.
What's overstatedMeta says full automation could arrive by the end of 2026, but the same agencies quoted in the same reporting say their own tests still show the AI creative tool performing worse on some tasks, and call that timeline unrealistic.
MetaTikTok
3 receipts
- Advantage+ already accounts for 60 to 70 percent of at least one agency's Meta ad spend.
Advantage+ campaigns now account for 60% to 70% of the agency's Meta spending
Marketing Brew, 7 Apr 2026 ↗ - The same agencies say their own A/B testing still shows the generative tool underperforming.
We do experiment on a regular basis with the pre-existing functionality and just consistently see worse results.
Marketing Brew, 7 Apr 2026 ↗ - Marketers remain hesitant to hand ad creative fully to the algorithm.
There's a lot of hesitation as to using GenAI specifically for ad creative right now.
Marketing Brew, 7 Apr 2026 ↗
- Advantage+ already accounts for 60 to 70 percent of at least one agency's Meta ad spend.
- 4
AI SDRs promise a fully autonomous rep, the receipts show a human-monitored draft machine
How it works. Clay, 11x and Artisan each sell some version of an AI worker that finds leads, enriches contact data, writes personalized emails and books meetings without a rep touching the sequence. Clay has become the data and enrichment backbone that many of these agents plug into, rather than a rep-replacement product itself.
Why it matters. Sales teams can run far more outbound volume with the same headcount, but a TechCrunch investigation found one leading vendor's own employees describing inflated revenue numbers and heavy churn behind the same pitch, so buyers now have to audit the metrics, not just the demo.
What's overstatedThe pitch is 'AI closes deals'; the receipts show AI drafting and sending, with a human still approving the sequence, or, in at least one competitor's design, still running the actual sales conversation.
Clay11xArtisanMonaco
3 receipts
- Clay's valuation tripled within a year on demand for its AI-powered lead enrichment layer.
Clay Labs Inc., a startup using artificial intelligence to make business-to-business sales teams more productive, today announced that it has raised $115 million in funding
SiliconANGLE, 9 Sept 2026 ↗ - A TechCrunch investigation found 11x employees describing heavy churn and inflated ARR behind its AI SDR product.
We were losing 70-80% of customers that came through the door.
TechCrunch, 24 Mar 2025 ↗ - Rival Monaco built its AI-native sales platform to explicitly avoid having an agent pose as a human rep.
Monaco does not have an agent pretending to be a sales rep trying to sell to the customer
TechCrunch, 11 Feb 2026 ↗
- Clay's valuation tripled within a year on demand for its AI-powered lead enrichment layer.
- 5
CRM incumbents and AI-native challengers both want to own the agent that acts on your customer data
How it works. Salesforce (Agentforce) and HubSpot (Breeze, Growth Context) are wiring their own agents directly into the CRM data and workflows they already sell. AI-native challengers like Attio and Rox AI pitch a CRM built around agents from day one, arguing an agent bolted onto a decade-old data model cannot reason about an account the way one built for it can.
Why it matters. The CRM renewal decision stops being about fields and pipelines and becomes about whose agent gets permission to act on the customer record, which raises switching costs and makes the winner's position much harder to dislodge later.
What's overstatedEvery vendor here describes agents that prospect, research and update records; none of the receipts we found describe an agent unsupervised at the point of an actual customer commitment.
SalesforceHubSpotAttioRox AIMonaco
3 receipts
- HubSpot built its own answer-engine-optimization and Growth Context layer directly into the CRM it already sells.
HubSpot AEO is a new solution for answer engine optimization that helps marketers tap into that opportunity by understanding, tracking, and optimizing how their business appears in answer engines like ChatGPT, Gemini and Perplexity.
HubSpot, 14 Apr 2026 ↗ - Attio raised $52M to build out its AI-native CRM.
we've raised a $52 million Series B to accelerate our mission of building the AI-native CRM
Attio, Aug 2025 ↗ - Rox AI hit a $1.2B valuation deploying agents that research prospects and update CRM records directly.
deploys hundreds of AI agents. These agents monitor existing accounts, research prospects, and update CRM software.
TechCrunch, 12 Mar 2026 ↗
- HubSpot built its own answer-engine-optimization and Growth Context layer directly into the CRM it already sells.
Who's doing it
10 companies to know, 7 with a round we can source. Newest first where dated.
Search these companies in the startup database ↗ · Explore all research tools
ArtisanAI SDR / outbound agents
Sells 'Ava,' an AI BDR meant to run the entire outbound motion, prospecting, writing, replying and booking, with a large built-in contact database.
Latest · 22 Sept 2026A September 2026 platform comparison confirmed Ava's prospecting database has grown past 300 million contacts. Breakout ↗
ProfoundAEO/GEO brand visibility analytics
Tracks and helps improve how a brand is cited inside ChatGPT, Gemini, Perplexity and Google's AI answers, the analytics layer for AEO/GEO.
Latest · 15 Sept 2026Profound raised $180M at a $1.8B valuation, jointly led by Sequoia Capital and Kleiner Perkins. SiliconANGLE ↗
$180M Series D, 15 Sept 2026 source ↗
Clayoutbound data enrichment and GTM orchestration
Data enrichment and workflow layer that sales and marketing teams plug over 200 external data sources into, then let AI agents research accounts and draft outreach.
Latest · 9 Sept 2026Clay's valuation jumped to $7.1B alongside its $115M Series D, roughly $2.1B higher than at the start of the year. SiliconANGLE ↗
$115M Series D, 9 Sept 2026 source ↗
Otterly.aiAEO/GEO monitoring
AI-search monitoring tool that tracks brand mentions and citations across ChatGPT, Perplexity and Google AI Overviews.
Latest · 4 Aug 2026Otterly.ai's own 2026 research found 15% of website traffic already comes from AI agents and bots, with ChatGPT driving over half of that. Otterly.ai ↗
Sierraenterprise customer support agents
Enterprise AI customer service agents built by former Salesforce co-CEO Bret Taylor and ex-Google exec Clay Bavor, deployed by large regulated companies for claims, refunds and account changes.
Latest · 4 May 2026Sierra, on the back of its $950M Series E, says it now claims more than 40% of the Fortune 50 as customers. TechCrunch ↗
$950M Series E, 4 May 2026 source ↗
Rox AIAI account management / CRM agents
Deploys agents that monitor accounts, research prospects and update CRM records for enterprise sales teams, sold as an alternative to bolting an agent onto an existing CRM.
Latest · 12 Mar 2026Rox hit a $1.2B valuation on continued demand for agents that research prospects and update CRM software. TechCrunch ↗
MonacoAI-native CRM plus prospecting database
AI-native CRM and prospecting database from ex-Founders Fund partner Sam Blond, deliberately built so agents draft outreach but a human still runs the sales conversation.
Latest · 11 Feb 2026Monaco came out of stealth with an AI-native CRM and a built-from-scratch prospecting database. TechCrunch ↗
$35M Seed + Series A, 11 Feb 2026 source ↗
Decagoncustomer support / concierge agents
AI 'concierge' agents for consumer-facing companies handling support conversations across chat, voice and email.
Latest · 27 Jan 2026Decagon added over 100 new enterprise customers, including Avis Budget Group and Block, alongside its Series D. Decagon ↗
$250M Series D, 27 Jan 2026 source ↗
AttioAI-native CRM
AI-native CRM built around a shared context layer, with agents that research accounts, answer questions and run multi-step workflows inside the CRM itself.
Latest · Aug 2025Attio raised a $52M Series B led by GV to build out its AI-native CRM. Attio ↗
$52M Series B, Aug 2025 source ↗
11xAI SDR / outbound agents
Sells 'Alice,' an AI SDR meant to prospect, message and book meetings across channels, the company that came to define both the AI-SDR pitch and its credibility problem.
Latest · 24 Mar 2025A TechCrunch investigation found 11x's own employees describing inflated ARR and 70-80% churn behind its AI SDR pitch. TechCrunch ↗
$50M Series B, 30 Sept 2024 source ↗
Your move
Our read, built on the receipts above.
The outbound AI-SDR layer is crowded and cooling after 11x's own numbers unraveled under reporting. The open lane is the boring reliability layer nobody has built yet, an independent auditor that checks an AI agent's claimed resolution rate, reply rate or churn against the vendor's raw logs, the way TechCrunch effectively did to 11x by hand.
Watch whether an outcome-based pricing model survives a second-year renewal at the same rate, that is the real test for the whole customer-support category. Red flag to probe in a pitch: a 'contracted ARR' number that counts trial customers who never converted past 90 days, the exact gap reporters found inside 11x's books.
Pull your own CRM or support tool's per-agent numbers this quarter, resolution rate, reply rate, meetings booked, and reconcile a sample of 20 by hand against the raw tickets or threads. If the dashboard and the manual count disagree by more than a few points, that gap is your real number, not the one in the vendor's deck.
Learn to interrogate an AI vendor's headline metric instead of accepting the dashboard: ask for ticket-level or contract-level data behind any published resolution or churn rate before signing a renewal. Start with whatever support or CRM tool your own company already has an active contract with.
What's next
- Dec 2026 Meta says full end-to-end automation of Advantage+ ad creation could arrive marketingbrew.com ↗
Other industries
- FinanceBanks move agentic AI from chatbots to running back-office and dealmaking work
- Software engineeringSpaceX buys the leading AI coding startup outright, for $60 billion
- HealthcareAmbient scribes go from pilot project to clinical infrastructure
- LegalLegal AI vendors go from research assistant to agent that produces the deliverable
- Real estateLeasing and resident calls move to voice AI agents that never sleep
- Jobs, HR and recruitingAI runs the first-round interview, not just the resume filter
- InsuranceClaims move from adjuster-first to AI-first, humans supervise the exceptions
- CybersecurityAutonomous AI agents take over penetration testing and vulnerability hunting
- Consulting and accountingAI-written client reports become a recurring, public Big Four embarrassment
- Retail and e-commerceShopping agents get a checkout button, and Amazon spends 2026 deciding who's allowed to press it
- EducationNYC and LA ban AI for students, not for teachers
- Media and creativeAI video splits into model labs that train and aggregators that resell
- Manufacturing and roboticsHumanoid robots move from demo livestreams to paid warehouse shifts
- Transport and robotaxisRobotaxis go from demo to metered fare across a dozen-plus US cities
- Energy and data-centre powerGrid rules get rewritten mid-buildout, state by state
- GovernmentGSA turns AI adoption into a government-wide subscription, then flips 'free' to metered
- Agriculture and foodAutonomous tractor fleets learn to run the tools behind them, not just the wheel