Daily enterprise tech brief ·
Enterprise AI releases: 23 August 2026
What changed. Why it matters to your business. What to test.
Lead signals
Research Hub ↗OpenAI cuts GPT-5.6 Sol pricing by over 20%
What's new: OpenAI has reduced GPT-5.6 Sol API and credit pricing for three months. Current API pricing is $5 per million input tokens and $30 per million output tokens.
Enterprises can test frontier intelligence on more complex production workloads while lowering the cost barrier for POCs and scaled deployments.
Source: OpenAI, 21 August 2026
Tricentis introduces readiness scoring for AI agents
What's new: AgentScore observes agents in real workflows, recommends evaluation criteria and generates composite scores with review, block or ship recommendations. Aida can autonomously explore applications and identify defects without an existing test suite.
Enterprises gain a structured way to determine whether an agent is production-ready instead of relying on successful demos or deterministic test cases.
Source: Tricentis, 21 August 2026
Epicor embeds vertical AI agents inside ERP
What's new: Epicor Prism is generally available across Latin America. Manufacturers and distributors can query live ERP data, analyze documents, create customizations and take pre-approved actions from Epicor Kinetic and Prophet 21.
Operational teams can investigate inventory, supply-chain and order issues without separate analytics tools or specialist report development.
Source: Epicor, 21 August 2026
UiPath improves human escalation from AI agents
What's new: Agent escalations can now be routed by workload, round-robin distribution, group availability or an LLM-selected recipient validated against the enterprise directory.
Organizations can scale human-in-the-loop agent workflows without creating manual assignment queues or repeatedly sending decisions to overloaded specialists.
Source: UiPath, 21 August 2026
Tamarac adds audited AI reporting for wealth managers
What's new: Tamarac's reporting beta can convert legacy PDF reports into modern templates and answer natural-language questions across client and account data. Every AI interaction is stored in immutable audit logs.
Wealth-management firms can modernize reporting libraries and generate client insights faster while retaining access controls and an auditable usage record.
Source: Tamarac, 21 August 2026
Serval launches an agent that builds IT automations
What's new: Catalyst is generally available. It examines ticket history and operating procedures, then drafts workflows, access policies and background agents that identify recurring IT issues before tickets are created.
IT teams can move from resolving repetitive tickets to proactively automating their root causes. More than 90% of Serval customers adopted Catalyst during beta.
Source: Serval, 20 August 2026
Executive summary & decision signals
Shortlist releases tied to an active business need. Verify primary sources and availability, then test business value in a bounded POC.
- Temporary frontier-model price reductions can widen the scope of production POCs.
- Agent evaluation is becoming a distinct production-readiness discipline.
- Enterprise agents are moving closer to governed actions inside core systems.
- Auditability and human escalation remain central to responsible automation.
All source evidence
Practical questions
What is the strongest pattern across these releases?
Enterprise agents are being measured, governed and connected to operational systems. Model capability still matters, but readiness scoring, audit trails, human escalation and cost control increasingly determine whether a deployment can scale.
Which update has the clearest near-term cost impact?
OpenAI has temporarily reduced GPT-5.6 Sol API and credit pricing, lowering the cost of testing frontier intelligence on complex professional workloads for the next three months.
Business implications are editorial assessments, not measured returns. Confirm release status, availability and eligibility before testing.
