Prepared for: investors evaluating Anthropic and OpenAI. Scope: US workplace AI use measured through the Recon Analytics AI Pulse survey.
A third pathway for growth
Two of the main growth pathways for AI companies like Anthropic and OpenAI include converting free-version users into paid-version users and increasing token consumption by developers within enterprises. Investors have only those two pathways to work with while the filings stay confidential: Anthropic submitted a draft registration statement on Form S-1 to the SEC on June 1, 2026, and OpenAI announced on June 8, 2026 that it had submitted one.
There exists a third pathway, one that Recon Analytics uncovered through our AI Pulse survey of 319,595 US respondents fielded between August 2025 and July 2026. Among US adults who use a frontier AI tool for work daily, 14.0% pay for the tool with personal money. At 14.0%, the self-funded group is nearly half the size of the 30.9% employer-funded group. The self-funded segment looks like consumer revenue but is performing work inside the enterprise. With a sales approach informed by the value and experience ratings those self-funded users provide, AI vendors like Anthropic and OpenAI can convert some of this segment’s revenue to employer-funded, contract-secured, more readily expandable enterprise revenue. Until they do so, this spend remains invisible to the employer and absent from the AI vendor’s enterprise sales forecast.
Converting free-version users into paid-version users may seem the more obvious path to growth, because free-version users are 55.1% of at-work daily users. The obvious path, however, leads to lighter, less valuable work. For example, only 30.4% of free-version users bring data analysis to their AI tools, while 43.2% of self-funded users use their AI for that high-value task.
The self-funded segment represents a growth opportunity for AI vendors
The employee paying for an AI tool with personal money has done much of the evaluation needed in advance of an enterprise sale. She tested the price against a personal budget and kept paying because the tool brought value to her work. Self-funded users rate value for what they pay at +23.7 and employer-funded users rate it +23.5. The employer is therefore being asked to fund a product its own employee has chosen and can be reasonably expected to remain satisfied with. Users who pay personally already see value in the tool. That gives the vendor evidence of employee demand before asking the employer to take over payment.
Two employer decisions stand between a self-funded user and a contract. Allowing the tool is a security and compliance decision an IT department makes, and funding it is a budget decision that comes more easily once the IT department has cleared the tool. Employees carried personal smartphones into workplaces that issued them feature phones through the late 2000s, and IT departments either supported the devices employees had chosen or bought a governed alternative. The same order holds in workplace AI: among daily workplace users on a free version, 48.7% say their employer supports or reimburses the tool they use, so employer permission runs well ahead of employer payment.
Converting self-funded users onto employer-funded contracts is only the first step toward growth; the next step is taking this more durable, contracted revenue and expanding it across work groups and with higher tier products. Durability comes from contracts that renew annually versus individual subscription accounts that may be canceled at any time. Expansion comes from better employer visibility on how the tool provides value to systems and processes followed by increases to seats and integration.
Claude’s positioning: large opportunity for a high-value product with some risks
Compared to ChatGPT, Claude has a higher percentage of its workplace users paying for the tool with personal money, 29.4% for Claude versus 14.7% for ChatGPT. Counting only the users whose subscription somebody pays for, 44.5% of Claude’s users pay for it themselves against 31.5% of ChatGPT’s. Therefore both the growth opportunity from converting to employer-paid and the risk from month-to-month individual contracts canceling are proportionally larger for Claude.
Paying users and free users grade their tools differently enough that only the paying users answer the question an employer is being asked, so the ratings here count paying users alone. Claude’s paying users rate Claude at +38.9 on making them more productive, against +29.4 among ChatGPT’s paying users, on a scale that runs from -100 to +100. That advantage sits in productivity rather than across the board. Across the seventeen at-work attributes Recon Analytics measures, paying Claude users rate Claude higher than paying ChatGPT users rate ChatGPT on ten and lower on seven, close to an even split.
Claude reaches the workplace through the employee rather than through the IT department. Technical roles account for 49.8% of Claude’s paying workplace users, against 38.9% for ChatGPT, 38.7% for Google Gemini and 36.5% for Microsoft Copilot. Among free-version users, 32.9% of those whose primary tool is Claude say their employer supports or reimburses Claude, against 51.2% of those whose primary tool is ChatGPT.
For Claude the harder of the two employer decisions is the one the IT department makes. Security and privacy is the attribute Claude’s paying users rate lowest, at -10.9 against -2.1 for ChatGPT’s paying users, and it is the attribute an IT department weighs when it decides whether to allow the tool at all. That rating is a gate in front of the contract rather than a cost inside it.
Conclusions for investors
An investor should price the conversion of self-funded users into employer-funded users as retention and expansion rather than as new revenue. Neither company has to build anything for that conversion to happen. An employee who pays for Claude with personal money can move that account into an employer’s Team or Enterprise organization, and Anthropic cancels the personal subscription and refunds the unused time. OpenAI treats an employee’s personal ChatGPT account the same way when it merges into a company workspace. The personal payment stops as the employer payment starts. The price of a personal subscription is nearly the same as the price of an enterprise subscription; the growth comes from expansion of seats and use once the employer contract enables that expansion.
Anthropic’s Claude has twice the conversion density of OpenAI’s ChatGPT on a much smaller base and still this conversion opportunity is a substantial addition to the more familiar free-to-paid conversion and API token growth pathways. Potential conversion of self-funded to employer-funded is one workplace pool where Claude leads OpenAI by two to one. The same concentration carries the larger downside, because an employer that standardizes on a rival removes the self-funded user rather than leaving the account in place.