Sales assets · Orbit vs Unity Catalog

Orbit vs Unity Catalog

Five structural weaknesses, AE objection-handling, discovery questions, and a five-minute Cost Anomaly Demo stage script. Source of record: docs/sales/unity-catalog-battlecard.md.

Five structural weaknesses

  1. 1. Multi-engine blind spot

    Orbit answer: Orbit enforces at the agent runtime boundary — Anthropic, OpenAI, Bedrock, custom stacks — regardless of substrate.[databricks-competitive-brief.md lines 27-31]

  2. 2. ABAC gap

    Orbit answer: Orbit pairs with UC / Collibra / Immuta for ABAC and adds runtime enforcement on top — one slice of catalog-vendor overlap reauthored as a runtime policy.[savings-formulas-source.md line 59]

  3. 3. Lack of upward reporting

    Orbit answer: Orbit's audit log is first-class data the customer owns — spend, denial rate, and over-budget alerts map to agent type and policy, not to model-quality drift.[databricks-competitive-brief.md lines 31-32 + 43]

  4. 4. Audit-trail gap

    Orbit answer: Audit log ships queryable in the customer's warehouse, exportable to their SIEM — not trapped in a vendor console behind a per-seat license.[databricks-competitive-brief.md line 43]

  5. 5. Migration cost

    Orbit answer: Orbit runs outside the warehouse. Orbit runs outside the lakehouse. Agent FinOps decoupled from the customer's compute-substrate decision — no compute-substrate hostage.[databricks-competitive-brief.md lines 40-42]

AE objection-handling

Prospect says Why they say it You say (one line) Don't say Source
I already pay for Unity Catalog. They've bought UC and assume it covers agent spend. UC governs data inside Databricks; Orbit governs spend at the agent runtime, which UC doesn't see. "UC is bad at this." (It isn't — it's a different layer.) [brief lines 36-43]
AI Gateway will catch up. Databricks roadmaps are tracking the same problem Orbit solves. AI Gateway rate-limits are lakehouse-only and not in any SLA. Orbit enforces across runtimes today. "Databricks is behind." (Sounds defensive.) [brief lines 27-32]
All our agents run on Databricks. The classic lock-in posture; they don't yet feel the cost. When the substrate changes — and it will — Orbit decouples Agent FinOps from that renegotiation. "Then you don't need us." [brief lines 38-43]
UC already covers row and column access. ABAC is real; they're not wrong about it. Conceded — Orbit pairs with your UC for ABAC and adds runtime spend controls on top. "Unity Catalog can't do ABAC." (It can.) [sales-competitive row 2]
We could build the audit ourselves. Often true; rarely durable once the fleet doubles. The audit log is the product — it's first-class data you own, not a notebook that ages out. "It's a lot of work." (Sounds like a threat.) [brief line 43]
Orbit just overlaps with Immuta / Collibra. Fears overlapping spend on top of catalog tooling. One runtime control plane replaces a 10% slice of catalog overlap; ABAC stays with the catalog vendor. "You should replace them." [formulas line 59]
We'll just throttle upstream. They believe billing alerts are enough. Billing alerts are after-the-fact; Orbit blocks inside the agent loop, before spend accrues. "You can't do that." [formulas line 61]

Discovery questions

  1. How many of your agents are running outside Databricks Model Serving today?
  2. What's the size of your annual Databricks platform spend — Low, Mid, High, or Strategic?
  3. Which catalog tooling runs alongside Unity Catalog right now — Collibra, Immuta, both, neither?
  4. Where do agent audit trails land today — Unity Catalog lineage, your warehouse, a vendor console, or nowhere?
  5. When you swap the underlying model on one of these agents, what happens to the existing spend limits?

5-minute demo stage cues

  1. Minute 1. Spend bar — open the live dashboard. Point at the spend bar showing the Finance Agent under its $50 hourly cap. Mention agent type, metric, and ceiling — "this is one of $N agents."
  2. Minute 2. Live audit log — walk the prospect down the most recent audit-log rows. Each row carries policy name, agent type, action taken, and observed value — no notebooks.
  3. Minute 3. Trip the cap — send one oversized request ($75 against the $50 finance-agent cap). Orbit returns `block` with `limit: 50` and `overage: 25`; the audit row writes inside the same loop. Show the request that would have proceeded and didn't.
  4. Minute 4. The Unity Catalog gap — pause on the audit row and ask, "Would Unity Catalog have stopped this?" Pause for the answer. Use the gap from `brief.md` lines 27-33 to remind the prospect UC is data-substrate governance, not runtime enforcement.
  5. Minute 5. Bring it home — recap: every minute the agent runs unenforced is spend accruing without a budget owner. Sub in the floor; aim next conversation at a 30-minute scoping call. Close on the contact form below so the AE has a name, work email, and rough tier to follow up on.