Oracle Digital Assistant pricing has been restructured three times in five years. The current model offers three commercial metrics — per-intent, per-named-user, and per-conversation — and Oracle's sales reps default to whichever metric maximises the deal value for the specific use case in front of them. The same chatbot deployment can carry three completely different price tags depending on which metric the ordering document lands on. A 5,000-user enterprise HR chatbot priced per-named-user costs $720,000 over three years. The same chatbot priced per-conversation, at a typical 12 conversations per user per year, costs $108,000 over three years — an 85% gap. Oracle Digital Assistant pricing is not a published rate card, it is a negotiation. This guide explains the three metrics, the formulas Oracle uses to translate one to another, and the two contractual traps that quietly inflate the bill at renewal.
Oracle Digital Assistant (ODA) is Oracle's conversational AI platform. It runs on OCI, plugs into Oracle Fusion Cloud Applications, Oracle E-Business Suite, ServiceNow, Salesforce and any system reachable over REST. The product is built around skill bots — focused, intent-driven assistants — and a Digital Assistant layer that routes user utterances across multiple skills. ODA has been GA since 2018 and is one of the few Oracle products that competes head-to-head with the hyperscalers in a non-database category, against Microsoft Copilot Studio, Amazon Lex, and Google Dialogflow CX.
The Oracle Digital Assistant pricing question matters because ODA is one of the few Oracle products where the commercial model has moved more than twice in recent memory. Buyers signing today inherit a far more sales-rep-discretionary pricing environment than buyers in 2020 did. Read the Oracle Cloud Licensing Guide for the broader OCI commercial pattern; this article zooms into ODA specifically.
Per-resolved-intent. Oracle counts each resolved intent — the bot correctly identified the user's goal and returned a response. Fallbacks and unresolved intents are excluded. Per-intent pricing is Oracle's default for variable-volume use cases such as public-facing customer service bots, where named-user counts are unknowable.
Per-named-user. Oracle counts the unique authenticated users that interact with the bot inside a billing period. This metric favours sparse internal use cases. The trap: Oracle's default contract clause defines named user as any user with the ability to interact with the bot, even if they never do. For HR bots deployed to all employees, the named-user count equals total headcount even if only 35% ever use it.
Per-conversation. Oracle counts a conversation as a continuous user interaction session, capped at a 30-minute idle timeout by default. The metric favours high-frequency use cases and is the cheapest of the three for internal helpdesks and IT-support bots. The trap: Oracle reserves the right to revise the conversation definition through unilateral exhibit update — historically, conversation timeout has moved from 60 to 30 minutes, doubling the billable count.
| Use case | Best ODA metric | Worst ODA metric |
|---|---|---|
| Public customer service bot | Per-resolved-intent | Per-named-user |
| Internal HR bot, 5,000 employees | Per-conversation | Per-named-user |
| Internal IT helpdesk, 24/7 | Per-conversation | Per-resolved-intent |
| Annual benefits enrolment | Per-conversation | Per-named-user |
| Sales lead-gen marketing bot | Per-resolved-intent | Per-named-user |
Oracle does not publish ODA list prices on oracle.com. The numbers below are the median negotiated rates we have observed across 40+ Oracle Digital Assistant contracts in 2024 and 2025. Treat them as benchmarks, not quotes.
To model the worked example from the lede: 5,000-employee HR bot at $4 per named user per month = $240,000/year = $720,000/three-year term. Same 5,000 employees averaging 12 conversations per year at $0.45 per conversation = $27,000/year = $81,000/three-year term. The 85% gap is real, and it is entirely a consequence of which metric the ordering document lands on. The Oracle Cloud Advisory service models this for every ODA deal we work.
Oracle's standard ODA exhibit defines named user as any individual authorised to access the digital assistant, whether or not they actually access it. The clause mirrors the long-standing Oracle Database Named User Plus definition, where the metric counts authorisation, not usage. For database, this is well understood by procurement teams. For ODA, it is almost never spotted in the redline.
The fix is a custom named-user definition in the ordering document. Push for active named users measured monthly, with a 60-day grace period for new hires. Oracle will resist; the fallback that we have closed on multiple deals is a named-user floor of 70% of total headcount with a true-up at renewal. The pattern is similar to the Java SE Universal Subscription Employee Metric — read the Oracle Java Licensing Guide for the parallel discussion.
Per-conversation pricing has a hidden risk that buyers rarely raise. The conversation metric depends entirely on how Oracle defines the boundary between two conversations. The current default is a 30-minute idle timeout. Oracle's standard exhibit reserves the right to update the definition through unilateral revision of the cloud-services policy URL.
This matters because Oracle has moved the timeout before. Customers signed on 60-minute timeouts in 2020 were rebilled at 30 minutes in 2022 — and many discovered the change only when the invoice arrived 1.4x higher than forecast. Defence: write the 30-minute timeout into the ordering document itself, with a clause that Oracle's right to revise the definition does not apply during the current term. The same pattern appears in Oracle's broader contract-revision tactics, which we cover in the negotiation guide.
For the broader pattern of Oracle commercial moves at renewal, see the Oracle negotiation guide and the support cost reduction guide. ODA negotiations land best when they ride on the back of a larger OCI or Fusion deal where Oracle's quarter-end discount floor is materially lower than the standalone ODA negotiation.
Independent, buyer-side analysis. Fixed-fee, 10 business day turnaround. Former Oracle insiders, 25+ years, $1.8B in Oracle spend advised.
Oracle Digital Assistant uses three metrics: per resolved intent, per named user, and per conversation. Pricing varies by metric, deployment scale and whether Oracle is bundling ODA into an OCI Universal Credits commit.
An intent is a discrete user goal the bot resolves, such as Reset password or Check leave balance. Oracle's per-intent metric counts resolved intents, not raw NLU requests; misclassifications and fallback responses are excluded but escalations to human agents are billable.
For high-frequency use cases such as IT helpdesk, per-conversation is usually cheapest. For sparse use cases such as annual benefits enrolment, per-named-user is expensive because each user logs in once or twice; per-conversation wins. For unbounded public bots, per-intent caps risk.
Yes. Oracle Digital Assistant consumption draws from Annual or Monthly Universal Credits and qualifies for Support Rewards offsets against on-premises Oracle Support fees.
No. The conversational layer is licensed under ODA. Generative AI completions, when ODA is wired to Oracle Generative AI Service, are billed separately at the GenAI Service per-million-tokens rate.
Twice a month. Oracle pricing moves, audit-defence tactics, GenAI Service rate changes. Written by former Oracle insiders.
No spam. Unsubscribe any time. Independent — not affiliated with Oracle Corporation.