No one can clearly define an AI agent. LinkedIn’s engineering boss says that’s okay

By John KellContributing Writer and author of CIO Intelligence
John KellContributing Writer and author of CIO Intelligence

    John Kell is a contributing writer for Fortune and author of Fortune’s CIO Intelligence newsletter.

    Mohak Shroff
    Mohak Shroff is senior vice president and head of engineering at professional networking platform LinkedIn.
    Courtesy of LinkedIn

    Mohak Shroff, the senior vice president and head of engineering at LinkedIn, acknowledges that everyone wants a crisp definition explaining what an AI agent is and what workplace tasks they can perform. But he encourages a broad and expansive view of the emerging technology. 

    “An AI agent is going to be many things,” says Shroff, who leads engineering, operations, IT, and security at the Microsoft subsidiary. “Let’s get specific about the task.” 

    Tech giants like Salesforce and ServiceNow, alongside with a number of startups, have been racing to launch AI agents, which use foundational models to complete complex, multi-step task (though humans frequently check the work before the agentic systems complete any assignments).

    For chief information officers, increasingly inundated with agentic AI pitches from vendors, the details and definitions of this technology are important. Already, skepticism has been building about overspending on too many off-the-shelf agentic systems, though surveys of technologists frequently show a majority intend to deploy at least a few AI agents this year.

    LinkedIn’s agentic pitch to customers began last fall with Hiring Assistant, which allows recruiters to upload job descriptions, notes, and job postings into the tool, which can translate that information into desired role qualifications and flag qualified candidates. Over time, the intention is that Hiring Assistant will become uniquely personalized as the agentic system ingests each recruiter’s feedback and preferences.

    Shroff says Hiring Assistant gives recruiters more time to talk to candidates and build a desirable talent pool for their companies or clients, and less time doing routine tasks like performing searches online and sending cold emails to lure talent, often at a low response rate.

    LinkedIn says that 40 of the 55 initial batch of charter customers of Hiring Assistant were surveyed (or 73%) to report that the tool saved them at least one hour when sourcing for a role. LinkedIn InMail acceptance rate of candidates was 36% when Hiring Assistant was enabled versus 28% for manually sourced candidates. Early customers using the tool include German conglomerate Siemens, telecommunications giant Verizon, and software provider SAP.

    In the future, Shroff envisions that agentic technology can help with job training and coaching. “These are all agents,” he says. “But the manifestation of them is slightly different in each case. One’s an assistant, one’s a coach, one’s a tool.”

    Shroff anticipates that for the foreseeable future, humans will remain in the loop and will oversee Hiring Assistant and other AI agent tools LinkedIn launches. He says the systems aren’t accurate enough to handle most complex tasks over any extended period of time. “In the event that the AI fails, which can happen, we believe far more in the infallibility of humans or the ability of humans to adapt,” Shroff says.

    But he anticipates that the type of tasks knowledge workers are responsible for will change quite drastically. “Our role will start to morph from being the doers of the work to the supervisors of the work,” Shroff says. External studies tend to agree. An estimated 15% of day-to-day tasks may be handled by agentic AI by as soon as 2028, according to research firm Gartner.

    Shroff likens this to his own career trajectory. He joined the professional networking platform in 2008 as a senior engineer but then began to serve as a manager by holding various director- and VP-level roles on his way to becoming an SVP in 2017. The chief technology officer, chief AI officer, chief information security officer, and VPs of developer productivity, productivity engineering, and product engineering all report to Shroff. That’s due to LinkedIn’s strategy that everything built for the company’s workforce is also intended to be offered externally to recruiters and job seekers.

    “All those things fit into that single ecosystem and separating any of them would actually feel broken,” explains Shroff.

    LinkedIn’s own internal workforce is also excited about the possibilities for agentic AI. The company recently hosted a week-long hackathon and many of the ideas floated were focused on agentic and generative AI. More than 10% of the company’s workforce participated in some capacity. Shroff says his colleagues are also encouraged to take LinkedIn’s learning courses to get up to speed on what agentic means for them but also, advises they keep a very, very open mind. 

    “It’s just ridiculously early,” says Shroff. “The only answer, when that’s the case, is just keep playing.”

    John Kell

    Send thoughts or suggestions to CIO Intelligence here.

    NEWS PACKETS

    OpenAI closes largest-ever tech funding round. ChatGPT maker OpenAI finalized a funding round that brings in $40 billion from SoftBank and other investors that values the AI startup at $300 billion, almost double the valuation when it last raised money in October. Reportedly, the deal will put some added pressure on OpenAI to complete a corporate restructuring by the end of 2025 that would give its for-profit arm independence from the nonprofit that currently governs the startup. If that restructuring isn’t completed by the end of the year, SoftBank would have the option to reduce its total contribution to $20 billion from $30 billion, Bloomberg reports, though OpenAI could seek additional investors to add to that sum.

    Anthropic, Databricks are teaming up for AI agents push. A new five-year, $100 million pact will make Anthropic’s Claude large language models available to businesses that store their data in Databricks’ platform, allowing those customers to build AI agents, an emerging and buzzy application of AI that can purportedly perform more complex workplace tasks with little to no human intervention. The sales teams at Anthropic and Databricks will also be allowed to sell each others’ products, the Wall Street Journal reports. Vendors have been abuzz pitching AI agents or tools that allow businesses to build their own agentic technology, but as this newsletter has reported previously, CIOs are moving judiciously and want to avoid having too many agents in the workplace.

    FBI reportedly probing cyberattack at Oracle. Bloomberg has reported that hackers broke into Oracle’s computer systems and stole patient data in an attempt to extort medical companies to pay ransoms and that the breach is being investigated by the Federal Bureau of Investigations. It is currently unknown how many patients’ records were taken and the total number of health-care providers impacted was also uncertain, but the news outlet has reported the software giant earlier this month alerted some customers that hackers accessed Oracle’s servers some time after January 22. Oracle also reportedly told customers that hackers accessed older Cerner servers, an electronic health record business that Oracle acquired for $28 billion in 2022.

    CoreWeave’s IPO seesaw. AI startup CoreWeave raised less money than initially planned in its IPO last week, and had a weak first two days of trading—only for shares to come back surging more than 40% on Tuesday. CoreWeave is the first “pure play” AI company to go public since the debut of ChatGPT in late 2022, but as Fortune explained its customer concentration (75% of 2024 revenue come from Microsoft and Nvidia) and high debt load have left some investors wary. If Tuesday’s rally is any indication though, Wall Street may be getting over its reservations.

    ADOPTION CURVE

    Companies expect to adopt agentic AI at a faster pace than the initial gen AI push. One out of every two companies say they are already leveraging AI agents and 94% anticipate they’ll adopt agentic AI more quickly than generative AI, according to a survey of 1,000 IT and business executives conducted by tech vendor PagerDuty.

    IT leaders were also overwhelmingly bullish on the return on investment expectations, with 62% of companies expecting more than a 100% return on investment from agentic AI. Leaders also hope to avoid some mistakes they made with initial gen AI projects before moving forward with AI agents: 41% say they rushed or insufficiently planned for gen AI. Other top regrets include spending too much (40%) and not setting up proper guidelines (38%).

    The survey compiled responses from four markets—the U.S., U.K., Australia, and Japan—and was backed by PagerDuty, a software-as-a-service incident management platform for IT departments.

    Courtesy of PagerDuty

    JOBS RADAR

    Hiring:

    - Major League Baseball is seeking a senior director of corporate infrastructure, based in New York City. Posted salary range: $240K-$275K/year.

    - Pressed Juicery is seeking a director of IT, based in Los Angeles. Posted salary range: $135K-$165K/year.

    - Walmart is seeking a director of technology operations, based in Bentonville, Arkansas. Posted salary range: $110K-$220K/year.

    - Illinois Department of Innovation & Technology is seeking a CIO, based in Springfield, Illinois. Posted salary range: $131K-$141K/year.

    Hired:

    The Hartford announced two executive appointments, including naming Shekar Pannala to serve as CIO and oversee technology, cybersecurity, infrastructure, and cloud modernization. Jeffery Hawkins was named chief data, AI, and operations officer. Pannala previously served as CIO of insurer Chubb and CTO of financial analytics provider S&P Global. Hawkins is a former CIO at healthcare giant CVS Health.

    Every Friday morning, the weekly Fortune 500 Power Moves column tracks Fortune 500 company C-suite shiftssee the most recent edition.

    - Etsy appointed Rafe Colburn as CTO, effective May 5, and reporting directly to CEO Josh Silverman. Colburn most recently served as chief technology and product officer of Depop, a London-based shopping site that Etsy acquired in 2021. Prior to Depop, Colburn spent nearly 10 years at Etsy, when he led the e-commerce company’s engineering teams. 

    - Pearson has expanded the responsibilities of CTO Dave Treat, adding leadership of the education company’s digital and technology operations alongside his current architectural and innovation responsibilities. CIO Marykay Wells, who led the digital and technology team, will leave Pearson. Both changes go into effect in April. Treat joined Pearson nine months ago and previously held senior roles at investment bank UBS and consultancies Deloitte and Accenture.

    - NVISIONx announced the appointment of Ben Rowe as CTO, joining the software-as-a-service cybersecurity company to oversee product innovation and apply new technologies like generative AI. Rowe most recently served as senior director of engineering at cloud-based video surveillance company Arcules, which recently merged with video technology provider Milestone Systems.

    - Auregen Biotherapeutics named Scott Lauder as CTO, after initially serving in the role on an interim basis, joining the clinical-stage biotechnology company after most recently serving as chief technical operations officer at biotech startup Laronde Bio.

    - Commvault announced the appointment of Bill O’Connell as chief security officer, joining the data management software provider after most recently serving as global head of cybersecurity and privacy at Swiss-based pharmaceutical giant Roche. O’Connell also spent nearly 12 years in security leadership roles at HR management software provider ADP.

    - US LBM promoted Jonathan Greene, currently chief digital officer, to the role of chief digital and technology officer. He succeeds Andrew Cambell, who served as CIO over the past two years for the distributor of building materials. Previously, Greene held leadership roles at furniture e-commerce retailer Wayfair and several executive roles with management consulting firm Bain Capital.

    - Noble Panacea announced the appointment of Benjie Limketkai as CTO, after previously working closely on skin-care brand formulations with founder and Nobel laureate J. Fraser Stoddart, who died late in 2024. Limketkai has previously held various research and development director roles, including serving as a lead scientist at Intel.

    - IEX Group named Steven Bonanno as CIO, joining the equities exchange after serving as CEO of data analytics firm BXS. Before that, Bonanno served as CIO at trading technology provider Dash Financial Technologies, and held senior roles at Bloomberg, Nasdaq, and Direct Edge.

    This is the web version of CIO Intelligence, a weekly newsletter on the tech, trends, and news IT leaders need to know. Sign up for free.