Interloom raises $16.5M to develop enterprise memory for AI agents

Interloom raised $16.5M in a seed round led by DN Capital, with Bek Ventures and existing investor Air Street Capital participating. The company provides an enterprise operations platform that captures undocumented expert workflows into a persistent "context graph" memory for AI agents to reduce knowledge loss and improve automation. Proceeds will be used to further develop the platform and expand enterprise AI and workflow automation capabilities.

Analysis

Interloom-style “memory layers” will disproportionately reward platforms that already own identity, workflows and entitlements inside large enterprises. Over the next 12–36 months pilots will prove ROI by cutting repeat human touches on complex cases — we expect initial automation lift in targeted processes of 20–40%, but full cross‑enterprise rollouts will be gated by integration and governance costs. That implies winners are not just point vendors but cloud/ERP/workflow incumbents that can bundle memory with billing and security.

Second‑order effects: as companies stitch operational memory into agents, headcount-driven spend in low‑value BPO and ticket triage should compress, rerouting incremental IT budgets toward data plumbing (graph storage, observability, fine‑grained access controls). This raises acquisition interest from hyperscalers and ERP vendors looking to internalize differentiated context graphs rather than pay recurring integration fees — expect an acceleration in M&A discussions in the 12–24 month window. Conversely, open-source LLMs + strong internal data governance could commoditize the memory layer unless vendors lock in through proprietary schema and ID‑level entitlements.

Principal risks: privacy/regulatory pushback and label quality. Memory that preserves past resolutions exposes provenance and liability risks (GDPR, CCPA, sectoral rules in healthcare/finance), which could force redesigns or slow enterprise procurement cycles. Watch for contract terms and SOC/ISO certifications as near‑term gating items; a single high‑profile data/accuracy incident could flip vendor selection dynamics within weeks, not years.

Key Decisions for Investors

• Long ServiceNow (NOW) — 6–18 month horizon. Rationale: owns workflow/identity surfaces enterprises prefer to layer memory on; buy on pullbacks or around product announcements. Risk: execution on cross‑sell; reward: 20–40% upside if adoption accelerates in ITSM and HR automation.
• Long UiPath (PATH) — 9–24 month LEAP calls or long equity. Rationale: RPA is the natural execution layer for memory‑driven agents; catalysts are enterprise pilot conversions. Risk: margin pressure and competition from cloud native automation; reward: convex upside via optionality on automation acceleration (target 2–3x on option exposure).
• Pair trade: Long Snowflake (SNOW) / Short Elastic (ESTC) — 6–12 month horizon. Rationale: centralized data platforms that monetize persistent graphs benefit; standalone search vendors face displacement. Risk: SNOW valuation already rich; reward: capture re‑rating if context graphs drive material new storage/compute spend.
• Short selective BPO/outsourcers (e.g., WNS, EXLS) — 12–24 months. Rationale: repetitive operational volumes are most at risk of replacement by memory‑augmented agents. Risk: firms can pivot to higher‑value services; reward: 20–40% downside potential if contract renewals shift to automation.