Hardest Part

Executing automated state-changing provisioning and deprovisioning actions across undocumented or brittle SaaS APIs without violating existing zero-trust security policies.

Min Viable Scope

Automate digital onboarding and offboarding triggered by a single HRIS platform for exactly three downstream apps like Google Workspace, Slack, and GitHub. Explicitly exclude physical hardware lifecycle management, software license negotiation, and shadow IT discovery.

Cold Start Problem

The product requires deep read and write access across dozens of enterprise tools before it demonstrates generalized utility. Break this by hard-coding workflows for the top five most common developer tools and targeting high-growth startups to nail the engineering onboarding bottleneck.

Time To First Value

1-2 weeks to map existing IT access patterns and configure necessary IAM permissions

Data Moat Available

true

Technical Difficulty

Moderate

Wedge

The initial beachhead targets electrical and HVAC contractors managing 50 to 150 field technicians. This niche experiences high tool rotation across dozens of concurrent job sites and relies heavily on expensive, specialized diagnostic gear. Once the agent secures the daily tool checkout workflow via SMS, it expands horizontally into auto-ordering daily consumables based on job progress, eventually managing full supply chain procurement for the firm.

Timing

The availability of robust multimodal vision models allows crews to check equipment in and out simply by taking a photo of their truck bed. Simultaneously, reliable reasoning models can parse unstructured SMS requests for materials and accurately map them against complex, unstructured supplier catalogs.

Why This ICP

Specialty contractors possess high-value, highly mobile assets and operate on tight schedules where a single missing tool delays an entire project. They feel acute financial pain from equipment leakage but remain small enough to adopt new operational systems without multi-year enterprise procurement cycles.

Size Of Prize

Approximately 180,000 mid-market specialty trade contractors in the US spend an average of $15,000 annually on dedicated inventory clerks and preventable tool replacement costs, yielding a total addressable prize of roughly $2.7B.

Gap Narrative

Mid-sized field operations run on chaotic group chats and whiteboards to track high-value equipment and consumables across multiple sites. Traditional inventory software requires manual data entry that field crews actively avoid, resulting in hoarded tools, lost assets, and stockouts. These companies require an active agent that fields natural language requests, reads photos of tool beds, and automatically reconciles inventory without a traditional user interface.

Defensibility

Defensibility builds through deep workflow lock-in as the agent becomes the sole operating interface between the field and the warehouse. Over time, the system accumulates a proprietary dataset of crew-specific consumption rates, tool lifecycles, and optimal local supplier routing, enabling predictive material staging that competitors cannot replicate on day one.

Why This Thesis

The Agent approach directly matches the reality of field work, where technicians reject traditional software interfaces and drop-down menus. An agent operating entirely over SMS, WhatsApp, and voice acts as a hidden service layer, delivering the operational control of an ERP without demanding behavioral changes from the frontline workforce.

SAM

~$1.5B-2B US and European enterprises actively scaling Generative AI programs

SOM

~$50M-150M realistic 3-year capture targeting early-adopter Fortune 1000 IT teams

TAM

~50k global enterprise IT departments × ~$80k-100k/yr AI management software spend ≈ ~$4B-5B

Growth Rate

~35-45%/yr, driven by rapid enterprise adoption of fragmented AI tools and escalating security risks from shadow AI

Paid Comparable Spend

~$150k-300k/yr in dedicated IT administrative labor, manual compliance audits, and repurposed generic SaaS management subscriptions

Kill Thresholds

• < 50% of known internal LLM traffic routed through the platform after 45 days
• Pilot-to-paid conversion rate on $80k/yr contracts < 20%
• Average implementation time > 21 days due to environment complexity
• D30 retention of IT admin dashboard users < 40%

Leading Metrics

• Percentage of raw API keys revoked and replaced by system credentials within 14 days
• Daily token volume routed through the centralized gateway
• Number of departmental rate limits and token budgets actively enforced
• Hours from deployment to the first automated detection of unauthorized shadow AI
• Weekly active users (WAU) among IT security and compliance staff auditing access logs

What Proves Right

IT security and operations teams deploy the gateway and route over 80% of internal LLM API traffic through the system within the first 30 days. Customers sign $80k annual contracts without demanding custom integrations for proprietary internal models. Daily active usage remains high among security admins who log in to audit access events, adjust token budgets, and block unauthorized shadow AI deployments.

What Proves Wrong

Development teams actively bypass the gateway to use raw API keys directly, leaving the platform with less than 20% of total enterprise token volume. IT buyers refuse standalone budgets, insisting on using basic controls native to AWS or Microsoft. Proof-of-concept deployments fail to convert because the automated discovery fails to detect custom-built agents running on local infrastructure.