Applying Anthropic's "Observed Exposure" methodology to 7 Go-To-Market roles. Task-level analysis. Statistical patterns. What's actually happening vs. what's theoretically possible.
On March 5, 2026, Anthropic published "Observed Exposure," a new measure that tracks what AI is actually automating in real workplaces. Not theoretical capability. Actual deployment. The methodology combines O*NET task databases, real Claude usage data, and weighting for automation vs. augmentation.
The headline: a massive gap exists between what AI can do and what it is doing. Computer programmers sit at 74.5% observed exposure. Customer service reps at 70.1%. Sales reps at 62.8%. Actual coverage remains a fraction of theoretical capability across most occupations.
This report takes that framework and applies it to 7 core GTM roles. We decomposed each role into its constituent tasks, scored theoretical LLM capability and observed deployment, weighted for automation vs. augmentation, and calculated a GTM-specific exposure score.
GTM is not one story. It is seven stories at different speeds. Some roles are 60%+ automated today. Others have enormous theoretical exposure and almost zero observed deployment. That gap is the strategic window. It is closing.
Anthropic's approach uses three data sources: the O*NET task database (~800 occupations), real Claude usage data from the Anthropic Economic Index, and theoretical exposure ratings from Eloundou et al. (2023). The key innovation: weighting by actual deployment, not just what is possible.
We adapted this framework for GTM by: (1) decomposing each role into 10-14 core tasks using O*NET classifications and GTM operational reality; (2) scoring theoretical LLM feasibility via the Eloundou framework; (3) estimating observed deployment from AI tool adoption patterns across GTM stacks; and (4) weighting by full automation (API/workflow) vs. augmentation (human-in-loop).
Note: Where Anthropic used proprietary Claude API traffic, we estimated observed deployment through market adoption data, GTM tool integrations, and practitioner surveys. Directionally consistent but different signal sources.
The big picture first. Theoretical exposure (what AI could do) vs. observed exposure (what AI is doing) across all seven GTM roles.
Blue = theoretical capability | Pink = observed deployment. The gap = unrealized automation potential.
RevOps has the highest observed exposure. It is the GTM role where AI is most actively deployed in production workflows right now. Marketing has the highest theoretical ceiling, but a significant gap to observed. Customer Success is the sleeper: high theoretical, low observed, massive opportunity window.
Larger gap = more unrealized AI potential. These roles will see the most change as deployment catches capability.
Task-level breakdown for each GTM role. Every score follows Anthropic's methodology: theoretical feasibility (can an LLM do this at 2x speed?), observed deployment (is it happening?), and automation vs. augmentation weighting. Click any role to expand.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Pipeline review & forecasting | 18% | 1.0 | Yes | Augment | Closing |
| Board & executive reporting | 12% | 1.0 | Yes | Augment | Closing |
| Cross-functional GTM strategy | 15% | 0.5 | No | - | Wide |
| Team coaching & 1:1s | 14% | 0 | No | - | N/A |
| Deal strategy & escalation | 12% | 0.5 | Partial | Augment | Moderate |
| Territory & segment planning | 10% | 1.0 | Partial | Augment | Moderate |
| Comp & incentive design | 6% | 0.5 | No | - | Wide |
| Hiring & org design | 8% | 0.5 | No | - | Wide |
| Customer/partner exec engagement | 5% | 0 | No | - | N/A |
Revenue Leaders sit at the intersection of strategy and judgment. AI theoretical capability is moderate here. Real-world deployment is minimal. The highest observed exposure is in data-heavy tasks: pipeline forecasting and board reporting, where AI augments rather than replaces.
The 34-point gap between theoretical (52%) and observed (18%) is the second widest of any GTM role. Anthropic found no systematic unemployment increase for highly exposed workers. For Revenue Leaders, the story is augmentation, not displacement. But the structural question is real: organizations are starting to ask whether they need one VP of Revenue Operations + AI instead of separate RevOps and strategy heads.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Account health scoring & monitoring | 15% | 1.0 | Yes | Automate | Closing |
| Onboarding playbook execution | 14% | 1.0 | Yes | Augment | Closing |
| Customer communication (QBRs, check-ins) | 16% | 1.0 | Partial | Augment | Moderate |
| Churn risk identification & intervention | 12% | 1.0 | Yes | Automate | Closing |
| Expansion/upsell identification | 10% | 1.0 | Partial | Augment | Moderate |
| Product feedback aggregation | 8% | 1.0 | Partial | Automate | Moderate |
| Strategic account planning | 10% | 0.5 | No | - | Wide |
| Relationship building & executive alignment | 10% | 0 | No | - | N/A |
| Internal escalation & advocacy | 5% | 0.5 | No | - | Wide |
Customer Success is the hidden disruptor. At 68% theoretical, it rivals Customer Service Reps (70.1% in Anthropic's data). But observed deployment at 29% reveals a 39-point gap, the widest in our analysis.
Anthropic found Customer Service Reps at 70.1% observed exposure. The leading automated task: "confer with customers to provide info, take orders, handle complaints." In CS, the same task exists, wrapped in more strategic context. Health scoring, churn prediction, and onboarding playbook execution are already automated via API. The relational tasks (executive alignment, strategic planning) remain human.
The hiring signal: Anthropic's statistically significant 14% drop in job finding rates for workers 22-25 maps directly to entry-level CSM roles (onboarding specialists, digital CS reps). These positions are most vulnerable.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| CRM data hygiene & enrichment | 14% | 1.0 | Yes | Automate | Closing |
| Reporting & dashboard creation | 16% | 1.0 | Yes | Automate | Closing |
| Lead routing & scoring logic | 10% | 1.0 | Yes | Automate | Closing |
| Forecasting model maintenance | 10% | 1.0 | Yes | Augment | Closing |
| Tech stack integration & automation | 12% | 1.0 | Yes | Augment | Closing |
| Process documentation & SOPs | 8% | 1.0 | Yes | Automate | Closing |
| Territory & quota modeling | 10% | 1.0 | Partial | Augment | Moderate |
| Cross-functional alignment & change management | 12% | 0 | No | - | N/A |
| GTM strategy & planning support | 8% | 0.5 | Partial | Augment | Moderate |
RevOps is the most exposed GTM role. Period. At 43% observed exposure, it approaches Anthropic's benchmark for "Computer User Support Specialists" (46.8%). The theoretical ceiling of 72% means we are past the halfway point.
This tracks with Anthropic's finding that "Computer & Math" and "Business & Finance" categories show the highest observed coverage. RevOps sits at the intersection of both. The dominant pattern is automation, not augmentation. CRM hygiene, lead routing, reporting, and SOP creation are increasingly handled via API, not human-in-loop.
Using Anthropic's regression: for every 10-point increase in observed exposure, BLS growth projections drop 0.6 points. RevOps is not disappearing. But the RevOps professional of 2028 will look more like an AI orchestrator than a Salesforce admin.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Sales content creation (decks, battlecards) | 18% | 1.0 | Yes | Augment | Closing |
| Call/demo coaching & feedback | 14% | 1.0 | Yes | Automate | Closing |
| Playbook & methodology documentation | 12% | 1.0 | Yes | Augment | Closing |
| Competitive intelligence gathering | 10% | 1.0 | Yes | Automate | Closing |
| Onboarding program design & delivery | 12% | 0.5 | Partial | Augment | Moderate |
| Training facilitation (live) | 10% | 0 | No | - | N/A |
| Tool & process adoption management | 8% | 0.5 | No | - | Wide |
| Cross-functional needs assessment | 8% | 0.5 | No | - | Wide |
| Performance analytics & skill gap analysis | 8% | 1.0 | Partial | Augment | Moderate |
Enablement is undergoing an identity shift. The content creation engine, historically the heaviest time allocation, is being automated fast. Battlecards, competitive briefs, and onboarding materials are among the highest-deployment AI use cases across the GTM stack.
Anthropic's "Market Research Analysts and Marketing Specialists" finding (64.8% observed exposure) is directly parallel. The leading automated task in that category: preparing reports and translating complex findings into written text. That is the same work Enablement does with sales content.
The wildcard: call coaching. AI tools (Gong, Chorus + AI summarization) have pushed coaching into "observed" territory via automation, not augmentation. AI does not help the coach coach better. It replaces the initial feedback loop entirely.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Account research & call prep | 12% | 1.0 | Yes | Automate | Closing |
| Email & proposal drafting | 14% | 1.0 | Yes | Augment | Closing |
| CRM updating & activity logging | 10% | 1.0 | Yes | Automate | Closing |
| Discovery calls & needs analysis | 16% | 0.5 | No | - | Wide |
| Demo/presentation delivery | 14% | 0 | No | - | N/A |
| Negotiation & objection handling | 10% | 0 | No | - | N/A |
| Internal deal strategy & coordination | 8% | 0.5 | No | - | Wide |
| Forecasting & pipeline management | 8% | 1.0 | Yes | Augment | Closing |
| Relationship nurturing & social selling | 8% | 0.5 | Partial | Augment | Moderate |
Anthropic placed "Sales Representatives, Wholesale and Manufacturing" at 62.8% observed exposure, higher than our AE estimate. The reason: that BLS category includes transactional sales (order-taking, price quoting) that are far more automatable than complex B2B enterprise sales.
For AEs, the story splits cleanly. Administrative and prep tasks (research, CRM, email, forecasting) are automating fast. The core of the role (discovery, demos, negotiation, relationship building) sits at 0 or 0.5 theoretical. AI does not replace what AEs do in the room.
If 36% of an AE's time gets automated, that is not displacement. It is capacity creation. The question: do you keep the same headcount and let them sell more, or do you need fewer AEs who each close more? Anthropic's hiring data (14% drop for 22-25 year olds) suggests organizations are already choosing the latter.
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Prospect research & list building | 18% | 1.0 | Yes | Automate | Closing |
| Cold email/message drafting | 20% | 1.0 | Yes | Automate | Closing |
| Sequence/cadence management | 10% | 1.0 | Yes | Automate | Closing |
| Lead qualification (BANT/MEDDIC scoring) | 12% | 1.0 | Partial | Augment | Moderate |
| Cold calling & live conversations | 16% | 0 | No | - | N/A |
| CRM updating & activity logging | 8% | 1.0 | Yes | Automate | Closing |
| Meeting scheduling & coordination | 6% | 1.0 | Yes | Automate | Closing |
| ICP & persona refinement | 5% | 0.5 | Partial | Augment | Moderate |
| Social selling & community engagement | 5% | 0.5 | Partial | Augment | Moderate |
BDR/SDR is the GTM role where Anthropic's young-worker finding hits hardest. This is overwhelmingly a 22-28 year old role. It is 38% observed exposure with the dominant mode being full automation, not augmentation. Prospect research, email drafting, sequence management, CRM updates, and scheduling are all being automated via API.
The math: 62% of a BDR's time allocation maps to tasks with observed AI automation. That does not mean 62% of BDRs disappear. It means 1 BDR + AI can do what 2-3 BDRs did in 2023.
Cold calling (16% of time, 0 theoretical exposure) is the human floor. As long as phone-based outreach remains viable, the BDR role has a floor. But the composition is shifting from "volume outreach machine" to "strategic conversation specialist who leverages AI for everything else."
| Core Task | Time Weight | Theoretical | Observed | Mode | Gap |
|---|---|---|---|---|---|
| Content creation (blog, social, email) | 20% | 1.0 | Yes | Augment | Closing |
| Campaign performance analytics | 12% | 1.0 | Yes | Augment | Closing |
| SEO & keyword research | 8% | 1.0 | Yes | Automate | Closing |
| Email campaign design & execution | 10% | 1.0 | Yes | Augment | Closing |
| Ad copy & creative brief writing | 8% | 1.0 | Yes | Augment | Closing |
| Market research & competitive analysis | 10% | 1.0 | Partial | Augment | Moderate |
| Campaign strategy & planning | 12% | 0.5 | No | - | Wide |
| Brand strategy & positioning | 8% | 0 | No | - | N/A |
| Event planning & execution | 6% | 0.5 | No | - | Wide |
| Stakeholder communication & reporting | 6% | 1.0 | Partial | Augment | Moderate |
Marketing has the highest theoretical exposure (74%) of any GTM role and the widest absolute gap to observed (38 points). Anthropic's data showed "Market Research Analysts and Marketing Specialists" at 64.8% observed exposure, already top 10 in the entire economy.
Our marketing estimate of 36% observed is lower because we include strategic and event-based tasks. For the content-and-analytics subset alone, we estimate 55-60% observed, right in line with Anthropic's numbers.
The dominant pattern is augmentation, not automation. Content writers use AI as a first-draft engine. Campaign managers use it for analytics interpretation. Anthropic's framework weights augmentation at 0.5x compared to full automation. If marketing shifts from augmentation to automation (end-to-end AI campaigns), the observed score could jump 15-20 points in a single quarter.
X = % of observed AI usage that is full automation. Y = overall observed exposure score. Upper-right = highest displacement risk.
Anthropic distinguishes between automated use patterns (AI replaces the task via API) and augmentative use (AI assists a human). In GTM, this split is the most important predictor of job impact. RevOps and BDR skew automation. Marketing and Sales skew augmentation. Automation reduces headcount. Augmentation increases output per head.
Anthropic found a statistically significant (p < 0.05, barely) 14% decline in job finding rates for workers aged 22-25 in exposed occupations. In GTM: BDR (entry-level by design), junior CSMs (digital/pooled CS), and associate-level marketing (content writers, campaign coordinators). If you are planning hiring for these roles at 2023 ratios, the data says you are over-hiring.
Theoretical exposure precedes observed exposure. The gap between them is unrealized automation potential. The roles with the widest gaps (Customer Success at 39 points, Marketing at 38 points) are the ones most likely to see rapid acceleration. Not the most exposed today. The most exposed tomorrow.
Applying Anthropic's regression coefficient (-0.6 percentage points in BLS growth per 10-point increase in observed exposure) to GTM: RevOps growth projections sit approximately 2.6 points below baseline. BDR approximately 2.3 points below. Not catastrophic yet. But the earliest measurable signal of a structural shift.
Anthropic references the economic literature on whether partial task automation leads to displacement or enhancement. In GTM, both are happening simultaneously in different roles. BDR tasks are modular enough that automating 60% genuinely reduces headcount needs. AE tasks are complementary enough that automating 36% makes the remaining 64% more valuable. Same technology. Opposite labor market effects.
| Rank | GTM Role | Theoretical | Observed | Gap | Primary Mode | Junior Hiring Risk |
|---|---|---|---|---|---|---|
| 1 | RevOps | 72% | 43% | 29 pp | Automation | High |
| 2 | BDR/SDR | 61% | 38% | 23 pp | Automation | Very High |
| 3 | Marketing | 74% | 36% | 38 pp | Augmentation | Moderate |
| 4 | Enablement | 65% | 35% | 30 pp | Mixed | Moderate |
| 5 | Customer Success | 68% | 29% | 39 pp | Shifting to Automation | High |
| 6 | Sales (AE) | 54% | 28% | 26 pp | Augmentation | Low |
| 7 | Revenue Leaders | 52% | 18% | 34 pp | Augmentation | N/A |
38% observed exposure. Automation-dominant. The BDR function is the GTM equivalent of Anthropic's "Data Entry Keyers" (67.1%). Plan for 40-50% fewer BDR headcount by 2028. Remaining roles shift to phone-first, relationship-led outbound.
39-point gap, the widest in GTM. The wave has not hit yet. Invest now in AI-native CS platforms. The window to restructure proactively is 12-18 months.
43% observed. Past the inflection point. The role is evolving from "systems administrator" to "AI workflow orchestrator." Hire for prompt engineering and AI workflow design, not Salesforce certifications.
AEs have the lowest observed exposure of any revenue-generating GTM role. Core selling skills (discovery, negotiation, relationship) score 0 theoretical. Double down on AE development. Redirect admin time savings into selling capacity.
Augmentation-dominant today. Could flip to automation overnight. The theoretical ceiling (74%) is the highest in GTM. One platform release enabling end-to-end campaign automation moves the observed score 15+ points in a quarter.
Anthropic's 14% drop in hiring for 22-25 year olds is already visible in GTM. Entry-level BDRs, content writers, digital CSMs at pre-2024 ratios? You are building a team structure that AI is actively dismantling.
AI does not overcome GTM roles. It overcomes GTM tasks. The roles that are 80%+ task-automatable get absorbed. The rest get restructured around the human-only remainder. That restructuring follows three distinct waves.
Every GTM role has a "from" and a "to." The transition is not replacement. It is redefinition. Each card shows what the role looks like today vs. where the data says it is heading, plus the human floor that AI cannot reach.
When AI makes average output free, exceptional output becomes the only thing worth paying for. The economics are simple: scarcity drives value. AI collapses the cost of execution to near zero. The things that cannot be replicated at zero marginal cost become exponentially more valuable.
There is a world where people choose their own path. It is already happening. Customers want AI for "check my balance" and want a human for "my mortgage has a problem." B2B is the same dynamic, amplified. A VP of Engineering evaluating a $200K platform purchase is not doing that through a chatbot. They want to read the room. They want judgment, not output.
Smart GTM organizations will segment, not force. High-velocity, lower-ACV motions go AI-first. Strategic, high-ACV, complex motions go human-first with AI behind the scenes. The companies that force one path lose whichever segment they ignore.
A Fortune 500 customer paying $1.2M a year does not want to be "managed" by an algorithm. They want a person who picks up the phone. The AI handles health scoring, QBR prep, churn prediction, and usage analytics behind the scenes. The human shows up with insight, not just information.
Designs, monitors, and optimizes AI workflows across the GTM stack. Decides which AI does what. Troubleshoots when the system breaks. Architects the next iteration.
Synthesizes ambiguous information. Manages competing stakeholders. Makes bets with incomplete data. AI makes the inputs better. The human makes the call.
Trust-based, high-stakes interaction. The buyer needs to look someone in the eye. Fewer of these people. Each more productive. Each more highly compensated.
Defines what the brand sounds like. Catches when AI output drifts off-strategy. Architects the content system. Taste is the asset, not production speed.
The new role. Sits between AI capability and business adoption. Part change management, part technical implementation, part training. Every org needs this. Almost nobody has it staffed.
Not every role balances the same way. Here is where the data says each role settles by 2029, based on observed trajectory, theoretical ceiling, and the human floor for each function.
The future of GTM is not AI vs. humans. It is a talent market repricing. The middle gets hollowed out. Average execution at average cost gets replaced by AI execution at near-zero cost. The top gets elevated. Exceptional judgment, creativity, and relationship skills command a premium they never commanded before. Hire fewer people. Hire better people. Pay them more. Give them AI leverage. The era of scaling GTM through headcount is ending. The era of scaling through talent density is starting.
Task decomposition: Anthropic uses O*NET's standardized task database for ~800 occupations. We decomposed GTM roles using O*NET classifications, job description analysis, and operational knowledge. GTM roles do not map 1:1 to BLS categories.
Observed deployment: Anthropic uses proprietary Claude API traffic. We estimated through market adoption data (vendor reports, G2/TrustRadius), GTM tool integration patterns, and practitioner surveys. Different signal sources, directionally consistent.
Automation vs. augmentation: Following Anthropic, fully automated implementations weight 1.0x, augmentative use weights 0.5x. Classification based on dominant deployment pattern (AI auto-updating CRM via webhook = automation; AI writing a first draft a human edits = augmentation).
Statistical significance: Anthropic's unemployment analysis uses difference-in-differences estimation with CPS data. Effects are "small and insignificant" for overall unemployment but "just barely statistically significant" for young worker hiring. Our GTM projections are extrapolations from their coefficients. Directional estimates, not causal claims.
Source: Massenkoff, M. and McCrory, P. "Labor market impacts of AI: A new measure and early evidence." Anthropic, March 5, 2026.