Agent Adoption by Job Function: Sales, Support, Finance, HR, and Engineering Data

What the Data Says

By April 2026 enterprise agent adoption is no longer uniformly distributed. Some job functions ship agents at scale; others are stuck at pilot. This piece compiles adoption rates, the workflows that work in each function, and where each one stalls.

The data sources: McKinsey State of AI 2026, Deloitte's Generative AI in the Enterprise survey, and a few large vendor case studies that have published numbers.

The Adoption Map

flowchart LR
    H[High adoption: 50%+ of teams] --> Eng[Engineering]
    H --> Sup[Customer Support]
    H --> Sales[Sales SDR/BDR]
    M[Mid adoption: 25-50%] --> Mar[Marketing]
    M --> Fin[Finance ops]
    M --> Hr[HR ops]
    L[Low adoption: under 25%] --> Leg[Legal]
    L --> Risk[Risk and compliance]
    L --> Exec[Executive admin]

Engineering

The highest-adoption function in 2026. The killer apps:

• Code completion and chat (Cursor, Claude Code, Windsurf)
• PR summarization and code review
• Test generation
• Documentation generation
• On-call assistance and incident summarization

Productivity uplifts measured are 10-30 percent for senior engineers, 30-60 percent for juniors. The variance is wide because measurement is hard.

Customer Support

Highest-ROI function for many companies. The 2026 deployments:

• Inbound voice agents handling routine inquiries
• Chat agents on website and in-app
• Agent assist (real-time suggestions for human agents)
• After-hours coverage
• Post-call summarization and disposition

ROI ranges: 30-70 percent labor cost reduction on automated traffic; CSAT typically flat or slightly up.

Sales (SDR/BDR)

The fastest-growing area in 2026. The deployments:

• Outbound email composition (with human review on send)
• Inbound lead qualification (chat or voice)
• Meeting prep and account research
• Forecast assistance
• Call recording analysis and coaching

Productivity uplift on SDR-level work: 2-3x measured at firms that have committed to the rollout.

Marketing

Solid adoption, but more on content production than campaign automation:

• Long-form content drafting (with human review)
• A/B test variant generation
• SEO and GEO optimization
• Personalization at scale (more bark than bite in many cases)
• Creative concept exploration

ROI ranges widely depending on whether AI replaces or augments existing teams.

Finance Operations

Steady adoption in transactional and reporting work:

• Invoice processing and matching
• Expense report review
• Variance analysis on close cycles
• Forecast scenario assistance
• Vendor onboarding triage

The reasons it lags engineering and support: tighter audit and accuracy requirements, slower legacy systems.

HR Operations

Mid adoption, growing fast:

• Recruiting (sourcing, screening, scheduling)
• Onboarding question answering
• Policy Q&A and benefits navigation
• Performance review summarization
• Internal mobility matching

The constraints are mostly compliance- and trust-driven; AI in HR has more political weight than other functions.

Legal

Low adoption (under 25 percent of teams), but growing. The deployments:

• Contract review and redlining
• Document discovery
• Research summarization
• Regulatory question answering

The constraint is liability — the malpractice question for legal AI is unsettled and risk-averse partners are slow to deploy.

Risk and Compliance

Low adoption. The use cases that have worked:

• Policy Q&A for employees
• Transaction monitoring assistance
• Suspicious-activity narrative drafting
• Audit response preparation

Constraint: same as legal, plus regulator watchfulness.

Executive Administration

Low adoption despite seeming like a good fit. The reasons: highly personalized work, the bar for failure is high (a missed meeting embarrasses an executive), and tooling has not caught up to the workflows.

What This Means for Vendors

flowchart TD
    F[High-adoption functions] --> Eng2[Engineering tools: mature, competitive]
    F --> Sup2[Support: mature, vendor-rich]
    F --> Sales2[Sales: rapidly maturing]
    M[Mid-adoption] --> Op[Big opportunity for vertical ISVs]
    L[Low-adoption] --> Spec[Specialist tools, slow sales cycles]

Vendors entering the engineering and support markets in 2026 face crowded, mature competition. Mid-adoption areas (finance ops, HR ops, marketing) are where vertical AI ISVs are still landing big logos. Low-adoption areas reward patience and specialization.

What This Means for Enterprise Buyers

If your function appears in "high adoption," you should be deploying — vendor maturity supports it. If "mid adoption," you have time to choose carefully but should not be at zero. If "low adoption," start with narrow internal pilots and learn before committing budget.

Sources

• "State of AI" McKinsey 2026 — https://www.mckinsey.com
• "Generative AI in the Enterprise" Deloitte — https://www2.deloitte.com
• "AI in customer service" Forrester 2026 — https://www.forrester.com
• "Generative AI productivity" Stanford-MIT — https://digitaleconomy.stanford.edu
• a16z enterprise AI report — https://a16z.com