Contents
The Agent-Readiness League Table by Industry
KanseiLink aggregates the results of agents actually hitting APIs/MCPs across 301+ Japanese SaaS/API services. Each service carries an agent_ready badge in three tiers: verified (proven at over 80% success), connectable (an API/MCP exists but is unproven), and info_only (no API). This time we ran across 8 industries and 100+ services to compare which sectors lead on agent-readiness and which lag.
The headline: only 5 services hold the verified badge. Across more than a hundred services, just five have been proven that agents can genuinely trust them.
| Industry | Verified count | Representative services (success / reports) | Readiness |
|---|---|---|---|
| Accounting | 2 | freee 90.1%/212, Money Forward 92.9%/42 | Leading |
| E-commerce | 1 | Shopify Japan 94.3%/53, Rakuten 50%/2 | Leading |
| Communication | 1 | Slack 91.2%/113, Chatwork 65.9%/123, LINE WORKS 20%/5 | Split |
| Project management | 1 | Backlog 90.1%/91, kintone 78.7%/61 | Mid |
| Payments | 0 | Stripe Japan 72.7%/33 | Mid |
| CRM | 0 | HubSpot 75%/4, Sansan 61.1%/36, Salesforce 42.9%/42 | Lagging |
| HR | 0 | KING OF TIME 65.7%/35, SmartHR 39.1%/92, Talentio 35%/20 | Lagging |
| Developer tools | 0 | GitHub (no report data), GitLab, etc. | Unmeasured |
What stands out is that the 5 verified services cluster in specific industries: two in accounting, one each in e-commerce, communication, and project management. HR, CRM, payments, and developer tools have zero. This is not chance — differences in data structure and regulatory environment by industry are what split the difficulty of agent-readiness.
Three Reasons Accounting Leads
Accounting is the only industry with two verified services. freee MCP posts a 90.1% success rate over 212 reports — the highest report volume in the entire KanseiLink dataset, and the service agents hit most often and most successfully. Money Forward Cloud is also high at 92.9%. Why has accounting alone pulled so far ahead?
- The data is structured. Accounts, journal entries, invoices. Accounting data is typed from the start, making it easy for agents to handle. Domains with little ambiguity have high first-attempt success.
- Accuracy demands drove up API quality. In a domain touching tax and audit, "roughly right" is unacceptable. Vendors were forced to build robust APIs and error handling early.
- freee prioritized readiness. freee published OAuth PKCE auth, five-domain integration, and structured docs early, spinning up the agent-development flywheel first. The 212 reports are the payoff of that early investment.
Accounting did not lead because it was technically special. It led because it prioritized readiness early. Structured data was a tailwind, but the decisive factor was the vendor's choice to "tidy up for agents." That is a move other industries can reproduce.
The Structural Reasons HR and CRM Lag
By contrast, HR and CRM hold zero verified badges — and even their market leaders post low success. SmartHR at 39.1% (92 reports), Salesforce Japan at 42.9% (42 reports): flagships of their respective markets, yet failing on more than one in two agent attempts.
HR: the wall of permission complexity and personal data
HR SaaS handles employees' personal information, payroll, and performance reviews. Permission models are finely segmented, and scope requirements differ per endpoint. In SmartHR's 92 reports, api_error (36) and auth_expired (10) stand out: auth expires easily and API errors are frequent. Talentio sits at 35%, with 9 search_miss errors — agents can't even find the service in the first place.
CRM: customization kills standardization
CRM is overrun with custom objects and custom fields that vary per company, so the standard API alone doesn't reach real-world use. Salesforce Japan shows 12 api_error and 11 search_miss errors, plus an average latency of 474ms — the slowest in the dataset, 2–3x the verified tier (128–216ms). Slowness invites timeouts and failures. Sansan does somewhat better at 61.1%, but still falls short of the verified threshold (80%).
SmartHR and Salesforce are excellent products for human users. But "easy for humans" and "easy for agents" are different metrics. Rich features and flexible customization delight humans yet register to agents as "unpredictable complexity" that drives up failure. Market share does not rescue success rate.
It Splits Within an Industry Too — The Communication Lesson
Beyond inter-industry gaps, success rates split widely within a single industry too. The starkest case is communication.
| Service | Success | Reports | Latency | Badge |
|---|---|---|---|---|
| Slack MCP | 91.2% | 113 | 163ms | verified |
| Chatwork MCP | 65.9% | 123 | — | connectable |
| LINE WORKS | 20.0% | 5 | 128ms | connectable |
All three serve the same purpose — team communication — with similar auth approaches. Yet success rates span 91%, 66%, and 20%, a gap of more than 4x. The interesting one is Chatwork: 123 reports, more than Slack, yet stuck at 66% — used a lot, failing a lot. LINE WORKS's failures are entirely search_miss: a discoverability problem where agents never find it at all.
The differentiators are clarity of the auth flow, straightforwardness of endpoint design, machine-readability of error messages, and discoverability. The same structure that splits success rates even under identical OAuth 2.0 plays out within an industry too.
What This Gap Means
Three numbers on the industry gap
In the agent economy, how fast an industry gets agent-ready becomes a battle for category leadership. In an industry like accounting, where several verified services line up, agents can confidently delegate that category's tasks. By contrast, industries with zero verified — HR, CRM — remain in a gray zone agents see as "usable, but not trustworthy."
For SaaS vendors this cuts both ways. A lagging industry is also open territory where no one has become the "verified king" yet. The first service to break the 80% wall in HR or CRM has a strong chance of monopolizing that category's agent demand.
- Get discovered via intent keywords — align metadata to agents' intent language like "manage attendance" or "log a deal." The first step to killing LINE WORKS-style
search_miss. - Raise first-attempt success — publish structured docs for auth, key endpoints, and known pitfalls. This was accounting's decisive move.
- Make errors machine-readable — return messages agents can repair themselves, not vague 500s. Turn failure into the next success.
- Cut latency — Salesforce's 474ms drives up failure. Aim for the verified tier's 130–220ms.
FAQ
Which industry has the highest agent-readiness?
Accounting SaaS. freee (90.1%, 212 reports) and Money Forward Cloud (92.9%, 42 reports) both hold the verified badge — the only one of 8 analyzed industries with two verified services. Structured accounting data, regulatory accuracy demands, and freee's early readiness work explain it.
Why are HR and CRM SaaS behind on agent-readiness?
Both have zero verified, with even market leaders posting low success (SmartHR 39.1%, Salesforce 42.9%). HR has complex permission models and handles personal data, producing many errors; CRM's per-company customization defeats standardization. Salesforce is the slowest in the dataset at 474ms, further depressing success.
What is the difference between "verified" and "connectable"?
Verified means proven through real agent usage at over 80% success; connectable means an API/MCP exists but is unproven. Of 100+ services, only 5 are verified (freee, Money Forward, Slack, Backlog, Shopify Japan). Being connectable is not the same as being trustworthy.
Why do success rates differ even within the same industry?
Communication is the clearest case: Slack 91.2%, Chatwork 65.9%, LINE WORKS 20.0% — a 4x-plus gap. Clarity of the auth flow, endpoint design, machine-readability of errors, and discoverability all differentiate. Even identical OAuth 2.0 produces divergent success.
What should SaaS vendors in lagging industries do?
The starting point is AEO work that moves a service from connectable to verified: (1) metadata tuned for intent-keyword discovery, (2) structured docs for auth, endpoints, and pitfalls, (3) repairable error messages. Accounting's lead was prioritization, not technology — HR and CRM can follow with the same playbook.
The figures in this article aggregate measured outcome reports collected by KanseiLink from agents (as of May 2026). The cross-industry analysis is based on search_services (limit 15 per industry) and get_insights measured values: freee 90.1%/212, Money Forward 92.9%/42, Slack 91.2%/113, Backlog 90.1%/91, Shopify Japan 94.3%/53 (verified); kintone 78.7%/61, HubSpot Japan 75.0%/4, Chatwork 65.9%/123, KING OF TIME 65.7%/35, Sansan 61.1%/36, Stripe Japan 72.7%/33, Salesforce Japan 42.9%/42 (474ms avg), SmartHR 39.1%/92 (337ms), Talentio 35.0%/20, LINE WORKS 20.0%/5 (128ms). GitHub had no report data. Report counts and success rates are snapshots at aggregation time and shift continuously with agent activity. confidence_score ranges 0.2–0.79; figures for low-report services are indicative. The "leading/lagging" labels and the structural interpretation of the industry gap are analytical views of observed data and do not guarantee future market trends. Check each get_insights for the latest values.