OpenAI could charge up to $20,000 (£15,425) per month for some of its more specialised AI agents, according to technology business title The Information.
It’s highest-priced agents will be able to support “PhD-level research”, meanwhile a $10,000-per-month (£7,715) version will be capable of performing software-development tasks. A “high-income knowledge worker” agent is also rumoured and could be priced at $2,000 (£1,543) per month, although this pricing is still unconfirmed.
The high price point “immediately raises a lot of questions about what it’s actually offering in return,” says Stephen Do, the founder of UpPromote, an affiliate marketing business. “AI is already deeply embedded in ecommerce, from customer service automation to predictive analytics, but at that price, it has to go far beyond the usual applications.”
Mohbeen Qureshi, the vice-president of growth at Oppizi, a New York marketing tech startup, adds: “$20,000 a month for an AI agent is serious money. That’s not just an expense – it’s a salary for a senior data scientist, a couple of solid engineers or an entire offshore team.”
With many PHD-level researchers being paid in the region of $20,000 per year, what capabilities would these agents need to have to justify the high price tag?
Automation on ‘another level’
A $20,000 agent should capable of handling complex, multi-layered tasks without requiring constant human oversight, says Do.
“That means generating optimised content in real-time, managing customer support with near-human accuracy and making intelligent marketing decisions that drive actual conversions,” he says. “AI should be analysing traffic sources, identifying the best-performing affiliates, and adjusting commission structures automatically based on performance. It should be able to track and predict purchasing behaviour so accurately that it improves sales without manual intervention.”
Anything less, he adds, just “wouldn’t justify the cost”.
‘No generic responses and no vague predictions’
GenAI platforms, as they exist today, are biased in favour of pleasing the user, often opting to respond with what the user wants to hear rather than giving the most accurate answer. Sometimes they even hallucinate, making up events or citations. At other times, they will return statements so broad they are functionally useless.
“A company spending that kind of money will expect an AI that understands their industry inside and out,” says Qureshi. “No generic responses and no vague predictions – it should adapt to specific business needs, analyse massive datasets with zero fluff and deliver insights that are actionable, not just interesting.”
That means a hedge fund will expect this agent to help them make markedly better trades and marketing teams will want campaigns that are optimised dynamically rather than just summarising performance reports.
Abilities far beyond a ‘glorified chatbot’
The $20,000 per month price tag might not be a problem – provided the cost is justified. But that means much more than just improving productivity, says Kristijan Salijević, who is CEO of the online retailer GameBoost. For that cost, he says an agent should be able to spot patterns and opportunities that human teams might miss.
“If it can identify shifts in user behaviour before they happen, optimise business strategies on the fly or automate complex workflows without constant oversight, then maybe it’s worth considering,” Salijević says. “But if it’s just a glorified chatbot that runs a little faster than existing AI tools, it’s a hard pass.” Such an agent would need to have a strong, measurable impact on revenue, as well as informing high-level decisions and predicting market trends.
Andreas Vermeulen is head of artificial intelligence at Avantra, a cloud software company. He says would consider trialling a $20,000 agent – but only if it solves specific, high-value business problems. And he would expect a very clear return on investment, with provable value of more than $200,000 (£154,250) per month before committing to ongoing costs.
“In the current economic climate, $20,000 per month is a high-risk investment,” he says. “It’s 10 junior people’s salaries, minimum. That’s a big ask.”
Agentic AI needs to reduce ‘operational drag’
GenAI systems currently require sharp, human interrogation. If an AI agent could truly take over complex decision-making, including analysing lengthy technical documents and handling compliance workflows, it would “be worth every dollar,” says Arne Helgesen, IT leader and tech manager at Sharecat Data Services, a software company.
“The real issue isn’t price, it’s whether this AI agent can truly reduce operational drag,” Helgesen adds. “Many GenAI tools still require heavy human oversight, making them glorified automation scripts rather than real intelligence.”
It’s got to be transparent and auditable
The trend towards explainability in AI exists for a reason: organisations need to know why their platforms came to a certain conclusion. Despite the ‘open’ in OpenAI’s name, the company has shied away from transparency in recent years.
That could be a problem for these AI agents, suggests Nirav Chheda, who is the cofounder and CEO at Bambi, a medical technology company.
“If it’s a black box that can’t explain its decisions, it’s useless,” he says. “In regulated industries, every action needs to be auditable. If an AI makes a scheduling error or misses a compliance requirement, who’s responsible?”
“If I can’t trust it to operate at a human level of accountability, I wouldn’t trial it. Also, if it requires constant babysitting or retraining, it’s just an expensive toy. Real business AI needs to work at scale, not just impress in a demo.”
OpenAI could charge up to $20,000 (£15,425) per month for some of its more specialised AI agents, according to technology business title The Information.
It's highest-priced agents will be able to support “PhD-level research”, meanwhile a $10,000-per-month (£7,715) version will be capable of performing software-development tasks. A “high-income knowledge worker” agent is also rumoured and could be priced at $2,000 (£1,543) per month, although this pricing is still unconfirmed.
The high price point “immediately raises a lot of questions about what it’s actually offering in return,” says Stephen Do, the founder of UpPromote, an affiliate marketing business. “AI is already deeply embedded in ecommerce, from customer service automation to predictive analytics, but at that price, it has to go far beyond the usual applications.”
