5 Ways To Successfully Integrate Ai Agents Into Your Workplace

Techa Tungateja / Getty Images

ZDNET's cardinal takeaways

• Using AIs efficaciously intends focusing connected guidance and governance.
• You'll besides request to guarantee you train up talent to negociate your AI agents.
• Track and trace outcomes to create nan cleanable tech-enabled activity assistant.

It's not different to perceive business leaders talk AI being utilized to augment alternatively than switch staff, but what does that caller moving narration mean successful practice?

Also: AI's biggest effect connected your workforce is still to travel - 3 ways to debar getting near behind

Five business leaders explicate really you tin move AI into a trusted squad member.

1. Put clear guidelines successful spot

Tim Chilton, managing advisor astatine Ordnance Survey, nan UK's nationalist mapping service, said his statement has rules for really professionals tin usage AI connected a day-to-day basis, peculiarly for generative technologies.

"Copilot is rolled retired crossed our organization," he said. "We person guidelines and training successful spot for really to usage it. I'm successful nan consultancy team, truthful it's successful our liking to afloat understand this technology."

Chilton said he often uses Copilot, pinch coding and investigation arsenic nan 2 superior usage cases. Other professionals are besides getting involved.

"Tools for illustration Copilot are proving useful for desk-based research," he said. "At Ordnance Survey, a ample proportionality of our staff, compared pinch different organizations, are progressive successful research, whether that's done doctorates, assemblage links, aliases commercialized links."

Also: 5 ways to beryllium a awesome AI supplier manager, according to business leaders

More generally, Chilton said nan rules and regulations OS has successful spot are helping unit research AI-enabled advances successful different areas.

"Our information subject capacity is besides investing heavy successful exemplary building to use america and our customers. We are utilizing these technologies to make nan information we are going to nutrient successful nan early and are changing really we seizure and quality-assure that information," he said.

"When it comes to geospatial information, we're putting a batch of clip and effort into really we usage Earth study data, which is information from satellites, to automatically deduce features connected nan ground, which will past beryllium validated and assured earlier being put into our models. We're doing that activity ourselves and for immoderate commercialized customers arsenic well."

2. Have a dialog pinch your AI

Snowflake co-founder and president Benoît Dageville said we're not yet astatine nan shape wherever AI acts autonomously successful a activity scenario, but we mightiness beryllium soon.

"The immense measurement will beryllium erstwhile AIs will really commencement to do things, specified arsenic looking astatine each your emails and replying for you. Then, you must person afloat trust, because you are not location to supervise."

However, Dageville besides issued a connection of warning, suggesting that conscionable because your AI tin activity independently doesn't mean you should fto it.

"Blind spot isn't awesome either," he said. "It's important to person big discussions pinch AI and understand wherever you request to push nan exertion and guideline it further."

Also: 5 tips for building instauration models for AI

Dageville emphasized nan value of method considerations and processes, specified arsenic governance, security, integrations, and information entree permissions.

"AI must not spot documents it doesn't person nan correct to spot and shouldn't show you thing that you're not expected to know. So, nan spot is besides coming from nan level and each this governance," he said.

"But astatine nan extremity of nan day, AI is for illustration a quality being. And arsenic overmuch arsenic you spot someone, you mightiness still perceive to what they're saying and say, 'Okay, possibly I half work together pinch what you're saying, truthful possibly I should inquire different question.' It's a dialog."

3. Develop mid-level talent

Rom Kosla, CIO astatine Hewlett Packard Enterprise (HPE), said he likes nan conception of copilots that enactment arsenic AI-powered assistants to professionals, and that's thing he's explored successful his organization.

"We were moving pinch a institution that looked astatine enabling inferior development, and past you person a elder developer assigning tasks to agents, and you tin delegate them tasks connected databases and possibly networks, and it comes backmost pinch different scenarios," he said.

"But you get to take arsenic nan elder developer -- you tin springiness them their activity backmost to support improving it: 'I request you to do this.' And what happens is you alteration nan elder developer to person much arms and legs, to extend."

Also: Most AI projects are abandoned - 5 ways to guarantee your information efforts succeed

However, conscionable because you've sewage AIs picking up nan slack for lower-level tasks doesn't mean nan domiciled of inferior developer goes away.

As my workfellow Sabrina Ortiz discovered, companies request humans successful nan loop to guarantee AIs tin beryllium trusted to activity effectively. Refining quality talent intends allowing inferior developers to study and move into higher-level positions.

"No 1 becomes a elder developer conscionable because they threat their fingers. They still request to ladder up. They must commencement arsenic a inferior developer. They must build up that skill. They request to understand nan types of answers," said Kosla.

"Think of nan apprenticeship exemplary -- conscionable because we person each these AI capabilities, you still request nan skills to understand each furniture of nan stack. That's not going to spell away."

4. Assess your agent's deliverables

Antony Hausdoerfer, group CIO astatine car breakdown master The AA, said that AI agents should beryllium tracked, traced, and assessed for illustration their quality counterparts.

"Like immoderate personnel of your team, you request to gauge whether aliases not you tin spot them, and that's mostly by looking astatine nan acquisition and assessing what they've done and whether aliases not they delivered nan outcome," he said.

"You request to ask, 'Did they present connected their commitments?' You request to commencement applying those assessments to nan agentic world and usage those arsenic your impervious points."

Also: 4 questions to inquire yourself earlier betting connected AI successful your business - and why

Hausdoerfer suggested that uncovering nan correct emerging exertion tin beryllium reliable because nan marketplace is flooded pinch options. Business leaders should deliberation cautiously earlier trusting an supplier arsenic a personnel of their team.

"There are galore different levels of AI now, and there's still a batch of sound successful nan system. There are truthful galore applications," he said.

"You've sewage nan business-centric systems that will amended your productivity, specified arsenic Copilot and immoderate of nan ChatGPT capabilities. But past you've sewage nan much bespoke applications, which will beryllium transformational from a business perspective, successful position of what AI delivers for customers."

5. Create nan cleanable intern

Vivek Bharadwaj, CIO astatine clothing shaper Happy Socks, suggested nan accelerated gait of alteration makes it difficult to cognize wherever to move adjacent erstwhile bringing AIs into your team.

"It's a brave caller world," he said. "There are truthful galore things happening. I judge that successful galore cases, moreover nan meaning of agentic AI is inconsistent."

Also: 4 ways to move AI into your business advantage

To counteract this unpredictability, Bharadwaj gave his meaning of agentic AI to thief level-set nan narration betwixt humans and emerging technologies successful nan caller workplace.

"Agentic AI is, successful my mind, your cleanable exertion intern. Rather than reasoning of agentic AI arsenic a wide superintelligence, I deliberation of it arsenic a individual intern you tin usage to automate circumstantial parts of your workflow," he said.

"The effective personification who leverages agentic AI is simply a systems thinker. They tin understand and break down their activity into discrete components, truthful they tin opportunity what constituent -- quality aliases agentic -- contributes to nan other."

Get nan morning's apical stories successful your inbox each time pinch our Tech Today newsletter.