The news
Three enterprise software vendors used October 7, 2026 to tackle the same problem from different angles: AI agents that give inconsistent answers, take wrong actions or drift after launch. Infor, Atlassian (TEAM) and Kore.ai each presented a plan for enterprise AI agent reliability, according to SiliconANGLE coverage and company announcements.
At its Infor Velocity Week event, Infor argued that generic agents are the root of the problem. Suresh Jayaraman, Infor's senior vice president of product management, told SiliconANGLE that generic agents cannot always give the same answer every time. Infor instead builds agents on top of its industry applications for sectors such as industrial manufacturing, aerospace and defense, automotive, and food and beverage. Its 2026.10 release adds security guardrails and scope controls, and some actions, such as moving a customer's orders between accounts, still need human approval.
Atlassian announced the Agentic Multiplayer Protocol, or AMP, at its Team '26 Europe conference in Amsterdam. According to the company, AMP lets agents work alongside people in Jira, Confluence and Loom, and gives each agent its own identity with defined authority and scope, set by administrators. Atlassian also introduced Rovo Work, a mode in which its Rovo assistant handles multi-step tasks that people review and approve, SiliconANGLE reported.
Atlassian said its MCP server, which connects AI tools to its products using the Model Context Protocol standard, handles more than 15 million tool calls a day, a 15-fold rise in six months, and has nearly 2 million monthly active users. Chief executive Mike Cannon-Brookes said in the release that the winners will get people and agents working together as one team.
Kore.ai launched Autoloop, which it describes as an engine that keeps tuning agents after they go live. Customers set goals such as task completion, rule adherence, cost and accuracy, and Autoloop tests the agent and rewrites the specific part of its instructions linked to a failure, SiliconANGLE reported. Kore.ai cited its own 2026 Agent Productivity Index, in which 79% of enterprises reported reversing agent actions. Autoloop is available to customers on its Artemis edition, the company said.
The numbers
- Enterprises reporting reversed agent actions (Kore.ai index)
- 79%
- Atlassian MCP daily tool calls
- 15 million+
- Atlassian MCP monthly active users
- Nearly 2 million
- Kore.ai Global 2000 customers (company figure)
- 500+
Why CEOs should care
For CIOs and heads of operations, the common thread is that autonomy without controls is not selling. Before approving an agent rollout, ask vendors three questions: what business context the agent is grounded in, which actions require a human sign-off, and how errors are detected and fixed once the agent is live. If a vendor cannot answer the last one, the cost of fixing failures falls on your team.
CISOs should focus on identity. Atlassian's move to give each agent its own identity, authority and scope mirrors how employees are managed. Security teams should insist that every agent has a named owner, least-privilege access and an audit trail, and that agents built on third-party platforms are visible in the same identity systems as staff.
For CFOs, Kore.ai's figure that 79% of enterprises have reversed agent actions is a reminder that agent savings come with rework costs. Business cases should include the cost of monitoring, human review and rollback, not just the labor the agent replaces.
The bigger picture
The three approaches point to a market moving from showing what agents can do toward proving they can be trusted. Industry-specific context, human checkpoints and continuous evaluation are becoming standard features, and they favor vendors that already hold customers' data and workflows, such as ERP and collaboration platforms.
What’s next
Watch for independent evidence that these controls reduce errors in production, and for pricing changes as vendors package reliability features into premium tiers. Buyers should test agents on their own data and edge cases before wide deployment.
What “Fact-checked” means
Fact-checking means testing a story’s facts against the evidence before it is published. This story went through at least two separate checks before this version was published.
- What we checked
- Its names, figures, dates, job titles, quotes and who said what were checked against the story’s sources, including its main source where it could be opened. The headline was checked for accuracy and overstatement.
- How
- A first check reviewed the whole story. If it passed, a second, skeptical check went back to the sources to look for mistakes in the most important facts. If a check flagged the story, it was edited to fix the problems found, and a separate re-check then reviewed the whole story again.
- Who
- The checks are made by our newsroom, as steps kept separate from the writing, under rules set by our editor, Hussein Mukhtar. A story the checks still flag is held for the editor, who decides whether it is fixed, published or dropped.
- If something is wrong
- “Fact-checked” does not mean error-free. If a material error is found after publication, we correct the story and add a note saying what changed. Report an error
Companies in this story
