Skip to content
Security & Trust Talking point

HPE engineering lead says firms start agents on low-risk jobs and delay money tasks

HPE's Sadeepa Wijesekara describes what she sees: agents begin on email and websites, and money and purchasing come later.

W
WebPulse Newsroom
AI-assisted · 2 min read
Share on X LinkedIn
HPE engineering lead says firms start agents on low-risk jobs and delay money tasks
In brief
  • HPE's Sadeepa Wijesekara said the companies she sees pick one or two low-risk uses for agents, get good at them, then add riskier work.
  • She said she has not seen a compliance standard for agents. She said HPE puts seven layers of security around the agents it rolls out.

Sadeepa Wijesekara is the engineering chief for HPE Private Cloud AI. On The TWIML AI Podcast, she said that in her experience companies start AI agents on low-risk work. Money and purchasing come later. She spoke with host Sam Charrington in episode #785, published on 17 November 2026.

What was said

Wijesekara said readiness depends on a company's size, its industry and the type of work. Small firms move fast, with few people to train and few systems to connect. She named that as one reason startups succeed.

In large enterprises, separate teams own separate pieces, and a security team joins in when new software arrives. She described website updates, email automation and marketing as easier and less risky. Automating financial transactions, purchasing or supply-chain decisions sits at the other end.

The pattern she described: companies pick one or two uses and go deep. They get good at them. Then they take on the riskier work once they feel comfortable, adding security controls along the way.

Separately, she said HPE puts seven layers of security around the agents it rolls out. She also said she has not seen a compliance standard for agents.

Why it matters

Elsewhere in the talk, Wijesekara warned: "Somebody could add some poison information into a ticket or email or a calendar invite." She also said an agent "just bypasses all of that." The excerpts do not show what "that" meant, so we do not tie either remark to the rollout order.

Our reading: this explains why the two kinds of work differ. An agent that reads tickets and email also reads text planted by outsiders. In marketing, the worst case may be a bad message. If the same agent can pay invoices or place orders, planted text could move money. Controls built for human staff may not hold when software does the work. A firm that moves first into money tasks may be testing its controls with real transactions. She did not say that.

For buyers, a vendor's success story in marketing email says little about readiness for payments or procurement. Ask which kind of work the example involved.

The other side

Wijesekara did not claim that early movers pay the price. She framed fast movement as a strength of small firms. The staged approach is her own observation of what she sees, not a measured survey.

Her view also comes from a hardware and platform vendor that works with federal and defense customers, a setting with a very high bar. Firms in other settings may behave differently.

She left the standards question open. She said she was curious whether listeners knew of an agent standard. She believed people are working on agent security guidance, but she could not point to a finished one.

Written by the WebPulse Newsroom with AI assistance, and checked by our editorial review: every quotation was verified against the recording's transcript. How we use AI.

The conversation this talking point comes from