Reasoning Engine

Conducting research…

Step 1 / 5

Discovering sources
Identified 6 candidate sources across 4 publication types.
Analyzing sources
Extracted atomic claims; scored credibility, recency, and bias on each.
Cross-referencing
Detected contradictions and reconciled overlapping claims.
Synthesizing findings
Compressed claim graph into structural themes and a working thesis.
Generating intelligence
Drafted executive brief, evidence map, risks, and recommendations.

Research Telemetry

live · demo

Reasoning

92/100

Confidence

86/100

Evidence

88/100

Depth

94/100

Diversity

82/100

Synthesized Answer · Deep Mode

Build a vertical agent thesis

Three structural reasons vertical wins in 2026: (1) eval coverage scales with task narrowness — a contract-redline agent can have 10,000 graded examples; a horizontal copilot cannot. (2) Integration moats compound — each new system of record (Workday, Epic, SAP) raises the wall around incumbents. (3) Buyer is identifiable: a vertical agent sells to a VP of Claims or Head of RevOps with a budget line, not to 'every knowledge worker.' The losing thesis is 'AGI replaces all white-collar work' — the winning thesis is '50 narrow products replace 50 narrow workflows.' Best risk/reward: regulated verticals (insurance, healthcare RCM, legal, financial back-office) where switching cost is high and eval discipline is rewarded.

Deep Analysis

86% confidence

The next 18 months will be defined by the agent reliability curve, not the model capability curve. Winners will own the eval/observability layer and a defensible workflow graph in a regulated vertical.

1 · Reliability is the new capability

Across SWE-bench Verified, GAIA, and WebArena, gains over 12 months are 80%+ attributable to better planning, verification, and tool fluency — not larger base models. Reliability scales sub-linearly with parameter count and super-linearly with eval coverage.

2 · Vertical beats horizontal on unit economics

Bounded task graphs make verification tractable. Vertical agents post 70%+ gross margins and 4–8 month payback; horizontal assistants are stuck at <40% margins and ambiguous ROI attribution.

3 · Computer-use is the integration unlock

Pixel/DOM-level agents bypass API gaps and let agents touch the long tail of legacy SaaS. Latency and reliability still gate production for high-stakes workflows.

4 · The platform layer is unsettled

Open frameworks (LangGraph, CrewAI) and native SDKs (OpenAI, Anthropic, Google) are converging. Expect consolidation around 2–3 winners by 2027; portability via MCP-style standards is the swing factor.

SWE-bench Verified — top model score over time

6-quarter trajectory

Agent benchmark scores by provider

benchmark composite (0–100)

Claude (Anthropic)

GPT (OpenAI)

Gemini (Google)

Llama 4 (Meta)

Qwen 3 (Alibaba)

Mistral Large

Closed Open

Agent deployments by vertical

share of measured value (%)

Software engineering32%

Sales & GTM ops21%

Customer support19%

Legal & compliance14%

Finance & ops14%

Contradictions detected

Claim

Benchmarks show 60%+ agent task completion (SWE-bench Verified).

Counter

Production deployments report 25–40% on novel tasks — Gartner attributes the gap to eval-set leakage and easier issue distribution.

Claim

Horizontal copilots will win on distribution (a16z).

Counter

Vertical agents are posting 3–5× the revenue per seat with lower churn (Gartner, internal surveys).

Key Points

Pick workflows: high labor cost + bounded eval + structured output
Eval coverage is the real moat, not model choice
Integration depth (Workday, Epic, SAP) compounds the moat
Vertical gross margin median: 72%; payback 4–8 months
Regulated verticals reward eval discipline → highest LTV

Knowledge Graph

10 nodes · 11 edges

topicconceptcompanyentity

Auto-generated Insights

Trend

Vertical agents are out-scaling horizontal copilots in revenue per seat by 3–5×.

Contradiction

Benchmarks show 60% task completion; production deployments report 25–40% — eval-set leakage suspected.

Finding

Computer-use APIs cut integration timelines from months to weeks for legacy systems.

Signal

Eval/observability startups (Braintrust, Langfuse, Arize) are seeing the fastest ARR growth in the stack.

Structured Data

Extracted from sources

Enterprises with agents in production

31%

18pp YoY

Gartner Q1 2026

SWE-bench Verified (top closed)

65%

22pp YoY

real GitHub issue resolution

Vertical agent gross margin

72%

9pp YoY

Open frameworks lead on portability; native SDKs lead on tool fidelity and latency. Convergence expected within 12 months.

Cred

Auth

Fresh

Rel

●Center

Strongevidence

Refine your research

Compare LangGraph vs native SDKs Build a vertical agent thesis Map the eval/observability stack Identify computer-use security risks Project agent ARR by vertical Generate executive summary

Demo mode · All sources, insights, and data are mock-generated for illustration.