On the same Columbus Day that Workday agreed to acquire C1.ai for agent authorization, Freshworks signed to acquire Torii, a SaaS management company that recently extended into AI consumption tracking. Financial terms were not disclosed. Closing is expected in Q4.
Workday's deal answers Is this agent allowed to act? Freshworks' deal answers a different set of questions: What did this agent cost, who authorized the spending, and what's the bill going to look like next month?
What Torii Actually Does
Torii launched in 2017 as a SaaS management platform. Tools in that category discover every application an organization uses, including the ones IT never approved, track who's using what, identify unused licenses, and automate onboarding and offboarding workflows. Torii's discovery architecture pulls from identity providers, SSO, finance systems, contract systems, browser activity, desktop tooling, HR data, and direct application APIs. Gartner has named it a Magic Quadrant Leader in SaaS Management three years running.
For this deal, the relevant part is what Torii built on top of that foundation.
AI consumption tracking. Torii's AI Management Platform, announced in May 2026, applies the SaaS management model to a new cost object: AI tokens. Traditional SaaS management tracks licenses, which are fixed-price seats that are either used or sitting idle. AI consumption generates variable cost per request, per model, per project, and the provider bills it as a single aggregate. Nobody in finance can tell you which team spent how much on which model, or for what purpose.
The AI module attributes tokens to departments, managers, and cost centers. It matches AI licenses against actual usage, discovers unsanctioned AI tools, flags cost spikes, and projects future spend from usage trends.
How the token telemetry works. For local developer tools (Claude Code, Gemini CLI, Codex CLI, GitHub Copilot Chat), Torii uses the OpenTelemetry instrumentation those tools already support natively. OpenTelemetry is an open standard for emitting usage and performance data. Installation scripts configure each tool to send signals to Torii's ingestion endpoint: user email, machine identifier, model, input and output token counts, errors, and rate-limit events. For Claude Code specifically, Torii can also capture tool names, MCP server usage, and spawned subagents.
Prompt content is not captured by default. The mechanism ties workstation and user identity to token counts, so finance can allocate costs to a team instead of staring at an aggregate provider invoice.
Coverage today is four documented developer tools. Torii's broader cross-vendor dashboard depends on direct integrations, and their coverage varies by provider.
The Gap This Fills
Freshservice, Freshworks' IT service management product, had been expanding from ticketing into a broader operations platform. Its May 2026 launch put IT service management, operations management, asset management, and enterprise service management on a shared data layer. The 2024 Device42 acquisition added infrastructure discovery and application dependency mapping. Freshservice already had its own SaaS management product, and its Freddy AI Agent Studio included governance features for Freshworks' own agents.
The missing piece was token-level consumption telemetry. Traditional IT asset management assumes the price is known at purchase and the open question is whether the asset gets used. AI consumption breaks that assumption. Cost is generated per request and varies by model and prompt complexity, and a single developer can run up thousands of dollars in tokens in a week without anyone in finance knowing until the invoice arrives.
Freshservice already owns the IT service desk, the asset inventory, and the SaaS subscription record, and it already has relationships with the IT leaders responsible for cost management. Torii adds the instrumentation to extend that cost management into the fastest-growing and least-visible category of IT spending.
What Remains Unproven
The product maturity signal is mixed. Torii's May announcement said the AI Management Platform was "available today." Its current product page labels it "early access." Nothing in the public record explains the discrepancy. Freshworks is acquiring a five-month-old product before its maturity has been independently validated.
Spending-limit enforcement mechanics are undocumented. Freshworks' announcement says Torii can "set limits." The public documentation shows no policy model, no approval flow, and no evidence that a limit actually stops model execution rather than generating an alert after the fact. Operationally, those are very different things. Until the enforcement mechanics are documented, treat "spending limits" as aspiration.
Token telemetry coverage is narrow. Four developer tools are documented. Enterprise AI consumption extends well beyond developer tooling — into customer service agents, marketing tools, analytics platforms, and internal copilots. Tracking Claude Code tokens on a developer's laptop is a long way from attributing all AI spending across the enterprise.
Integration architecture is unpublished. As with Workday–C1, no public design shows whether Torii will remain a distinct product, feed Freshservice's configuration management database, or become a native module. Torii's CEO said the companies remain operationally separate until closing.
Maturity
Torii has raised $65 million across three rounds, including a $10 million Series A in 2021 and a $50 million Series B led by Tiger Global in February 2022.
The path wasn't smooth. Calcalist reported that Torii cut 28 of 95 employees in June 2023, roughly 30% of its workforce, citing revenue below expectations. LinkedIn currently shows 156 associated profiles, which suggests the company rebuilt headcount after the contraction, though profile counts are a loose proxy. Named customers include Hearst, Groupon, Koch, Instacart, and Palo Alto Networks. Torii does not disclose how many of them use the AI management module versus the established SaaS management product.
Freshworks is buying established SaaS management distribution and a new AI module whose production evidence is thin.
Two Theories of Agent Governance
Workday's C1.ai acquisition assumes identity governance is the foundational layer. Authorization happens at the point of execution, and in that moment it's binary: allow or block. As I argued in the companion piece, it's also architecturally upstream. Cost tracking, compliance reporting, and performance monitoring all depend on knowing which agents exist and what they're authorized to do. You can't attribute costs to an agent you haven't identified, and you can't enforce spending limits without the layer that controls execution.
Freshworks' Torii acquisition assumes something else. Most organizations don't yet have mature authorization frameworks for agents, but their AI costs are already growing faster than their ability to track them. The accounting surface sits at the point of visibility. It doesn't stop an agent from acting. It tells you what happened, what it cost, and whether the trajectory is sustainable.
On timing, Freshworks has the advantage. Right now, most enterprises run agents on inherited human credentials and ungoverned MCP connections, with no centralized identity management. Building a mature authorization framework takes policy design, connector deployment, and security team buy-in. The AI invoices keep arriving monthly in the meantime. A CFO who can't tell the board how much the company spends on AI has a problem that consumption tracking can address this quarter.
On durability, Workday has the advantage. Cost visibility is valuable but likely commoditizable, since every cloud provider, AI platform, and finance tool has an incentive to offer better cost reporting. Identity governance is harder to replicate because it requires deep integration with every system an agent touches, and it gets stickier as more agents and policies are registered. An authorization surface accumulates institutional knowledge (approval chains, exception records, policy history) in a way a cost dashboard does not.
| Workday–C1.ai (Authorization) | Freshworks–Torii (Accounting) | |
|---|---|---|
| Control point | Agent identity and tool-call enforcement | Consumption tracking and cost attribution |
| Governs | Is this agent allowed to act? | What did this agent cost? |
| Timing advantage | Requires organizational maturity to deploy | Addresses acute cost-visibility gap now |
| Durability | High — accumulates institutional policy | Lower — cost dashboards are replicable |
| Buyer relationship | HR and finance leaders | IT leaders |
| Dependency position | Upstream — accounting depends on identity | Downstream — requires identity to mature |
Each acquirer already has some capability in the other's lane. Workday's Agent System of Record included cost budgeting and forecasting, and Torii already sells identity governance features. The line between these deals is useful for analysis and porous in practice.
What the Platform Positions Reveal
Each company is acquiring the governance capability that fits its existing data model and customer relationship. Workday owns the employee record, the organizational hierarchy, and the role structure, so adding agent identities and cross-system authorization is a natural expansion. Freshworks owns the IT service desk, the asset inventory, and the SaaS subscription record, so adding token-level AI consumption is the natural expansion.
In both cases, the platform each company already operates shaped which governance layer it could credibly build. Workday sells to HR and finance leaders who think in roles, permissions, and organizational structure. Freshworks sells to IT leaders who think in assets, costs, and operational visibility. The governance question each company chose to answer is the one that feels most urgent to its buyer.
If I had to rank these bets, Workday's is the stronger long-term position. Authorization is harder to replicate, stickier once deployed, and upstream of accounting in the dependency chain. Freshworks may see faster adoption, though. The cost problem is already acute, and the buyer — an IT leader managing a ballooning AI budget — already knows they need help.
Watch for post-close integration announcements. Both deals are signed; neither has closed. Whether these acquisitions produce unified governance surfaces or stay co-branded tools bolted onto existing platforms will determine which thesis actually pays off.
