Your next CEO could be an AI agent. That’s not to say it will be, but US-based R&D outfit Sente Labs has released Open Executive, an open-source AI project that can automate C-suite executive roles. Sente co-founder Alex Cohen explains that the company’s goal “is to transform management and leadership through the power of AI,” and said that Open Executive’s customisability makes it suitable for a wide variety of organisations.
While acknowledging that legal and regulatory issues may prevent corporations from appointing Open Executive AIs as their CEOs anytime soon, Cohen nonetheless predicts that businesses will become increasingly comfortable over time with assigning responsibility to Open Executive-based agents. And even though researchers are skeptical that we’ll see organisations hand over full executive control to AIs in the near future, they do affirm that executives may increasingly delegate many tasks and supportive functions to agents.
Three levels of customisability
Cohen says that Open Executive provides a “highly customisable” framework, enabling organisations to modify its default personas to the specific roles and functions they require, and to provide agents with differing levels of authority.
“Open Executive is persona-driven, so it is straightforward to configure it to replicate any leadership or management role,” says Cohen. “For example, Head of Marketing, Director of Software Engineering, or even a Project Manager.”
Users can deploy agents at three levels of authority. The highest level is fully autonomous, which organisations will use if they want to completely replace an executive with an agent. The next, intermediate level, involves duplication, creating a double for an executive that syncs with their personal data and apps (e.g. email, Microsoft Teams, Jira) to answer queries and contribute to workflows. Lastly, there is an “executive assistant” level of autonomy, in which an agent primarily handles administrative tasks such as scheduling, coordination and routine emails.
“We recognize that autonomous AI leadership is a novel concept,” Cohen says, “so flexibility was essential to allow organizations to adopt a comfort level that suits them. Over time, we anticipate that organisations will grant the system increasing autonomy, though proper initial configuration remains critical.”
Given the customisability of Open Executive, Cohen expects that companies will engage with Sente Labs to assist with installation and configuration, although they are free to deploy Open Executive independently. Either way, Cohen suspects that many organisations will take the plunge, given the various advantages it offers.
“First, by connecting natively to platforms like Salesforce, Microsoft Teams, Jira, and Confluence, it processes and analyses vast amounts of real-time operational data,” he explains. “Far beyond what an individual human executive could track. It leverages this comprehensive dataset to drive consistent, data-informed decisions.”
Sente Labs engineered Open Executive for “explicit” ethical alignment and professionalism, training its model on established accounting practices, business ethics and regulations, and case studies. This potentially makes it more ethical than at least some CEOs, while Cohen also argues that it can make more objective decisions.
“While AI-driven leadership is uncommon today, embedding explicitly programmed ethical guardrails into executive decision-making may soon become standard practice, much like adhering to GAAP principles,” he says. “Furthermore, an AI executive operates without ego and with significantly reduced personal bias, allowing it to make key operational decisions based strictly on objective evidence.”
There is a concession, however, that “legal and regulatory considerations” may prevent organisations from enabling agents to act as final decision-makers. Similarly, Cohen acknowledges that AI tools lack human presence, in that they’re unable to interact in a traditional, embodied sense and build strong interpersonal relationships.
On the other hand, Open Executive has one human quality: the ability to “dream.” This amounts to the capacity for some degree of self-reflection, whereby agents regularly analyse their existing plans, while similarly identifying and working to fill gaps in its knowledge sets.
“For example, it might evaluate our current product offerings, recognise a lack of competitive intelligence, and autonomously research competitors across the web,” Cohen says. “Once it processes that information, it integrates the findings directly into its own long-term memory.”
Ambiguity, trust and relationships
While Sente Labs is confident in the ability of Open Executive to replace or fill in for CEOs, outside experts have more sober opinions about agentic executives. For Nicos Savva, a professor of management science at the London Business School, the subjectivity inherent to C-suite roles makes them less amenable to full automation.
“I think this is VERY unlikely with current (or foreseeable) technology,” he says, explaining his skepticism in terms of the “messiness” of the tasks and responsibilities executives generally have, in that such functions cannot be explicitly articulated or codified.
“That’s not to say that AI will not change how these tasks are performed (or indeed who gets to perform them),” he says, “but for these tasks we are not talking about automation so much as augmentation.”
This kind of argument crops up repeatedly. According to Imperial Business School’s Dr Ying-Ying Hsieh, AI is likely to serve C-suite executives increasingly in a supportive capacity, enabling them to arrive at more informed decisions. That said, they won’t replace executives, who do considerably more than optimise decisions.
“They interpret ambiguous situations, negotiate competing interests, and build trust,” Hsieh explains. “They also motivate people, exercise judgement under uncertainty, and take responsibility for decisions whose consequences may be difficult to quantify.”
The element of trust and relationship-building is something that managerial experts highlight time and again. Generally, this element requires physical embodiment, a figurehead that employees and investors can look to, and around which an organisation can organise itself.
“People clearly do trust systems in plenty of contexts,” says Pedro Varela, the Head of AI at consultancy Slalom. “It’s just that trust in executive leadership specifically involves accountability and relationships, with real stakes attached to a specific responsible party.”
Who is responsible for the actions of an agentic CEO?
Accountability is the other serious issue, and it currently lacks a clear answer given the novelty of the technology.
“Who can a regulator or court actually hold to account when something goes wrong?” Verlea asks, adding that just because an organisation has given an AI system the authority to act, it doesn’t mean the same system assumes legal responsibility for the consequences of its actions.
And even if a company implements an agentic executive in a fully compliant way, the recent spate of rogue agent incidents would suggest there may be safety and security issues.
“Being compliant isn’t the same thing as being safe,” Verlea says. “You can satisfy every governance requirement and still be exposed to real harm.”
Warnings aside, there is a real consensus that executives will make increasing use of AI in the coming years. Not only that, but some experts suggest that CEOs could gain a meaningful edge by incorporating agentic AI into their work processes.
“A CEO with sound judgement and expertise can direct agentic AI to process data, stress test their ideas, and even prototype creative solutions at an unprecedented speed and scale,” says Professor Hatim A. Rahman of the Kellogg School of Management at Northwestern University. “These are significant capabilities at the C-suite level.”
The post This company wants to replace your CEO with an AI agent appeared first on Raconteur.