The axis of competition in AI is quietly shifting. For years it was all about "who builds the better model," but now the contest has moved to "how you embed that model inside a customer's organization." Model performance has leveled up across the board, and most of what's left to solve lives in deployment and operations. Where the data sits, how the work actually flows, who makes the decisions — an AI that doesn't know this context is only smart in the demo.
The first company to really grasp this was Palantir. Since the 2000s, Palantir has run an organization called FDE (Forward Deployed Engineer). Instead of selling a product and handing over a manual, they place engineers inside the customer's walls to solve real problems on the ground. In 2026, Big Tech is pulling that old idea back out and rebuilding it around AI.
ONTLE is the Korean version of that experiment. For each customer project, it "dispatches" an AI PM with a name and a title to handle everything from intake to progress management. So what do these dispatch-style services overseas actually look like, and where does ONTLE stand among them? Here we lay out the competitive landscape as the ONTLE team mapped it while surveying the market, broken down by type. Our read on these external services is anchored to their official announcements — and where a number can't be verified, we don't force it into place.
It's true that in the first half of 2026, Big Tech firms announced, one after another, companies built to "embed AI inside the customer." But those announcements weren't bunched into a single month, and they weren't a single event.
The three announcements are spread across two months, from early May to early July. The phrase "within a single month" doesn't hold up.
The numbers call for even more caution. The investment figures the press cites vary from source to source, and above all, an announced or raised amount is not the same as money actually deployed. "Announced the founding" and "raising X" describe a plan — not proof that the same sum has already been poured into customers. The moment you add up several companies' announced figures and write "tens of billions of dollars deployed in a single month," you've conflated plans with results. The direction — that the stakes are rising — is clear. But it's more honest not to treat these as confirmed deployment figures.
The same goes for sweeping claims about target customers. At minimum, Anthropic itself said it is aiming at mid-sized companies. Stretch that into "they focus only on big enterprises, finance, and the public sector" or "startups were never even an option," and you're tacking on a story the source doesn't support. It's natural for Big Tech's dispatch-style arms to be strong at large contracts — but no official material guarantees the sentence "they won't even deal with small and mid-sized businesses." A tendency for economies of scale to tilt toward big contracts is one thing; an official statement excluding a particular size is another.
The second type doesn't dispatch people — it sells, as SaaS, an AI that takes over an entire job function.
Ema offers role-based "AI employees" as a service for things like customer support, HR, finance, and sales. It leans toward pricing closer to the work actually done than to the number of seats. Devin (Cognition AI) is an AI engineer that autonomously writes code, debugs, and deploys. It started at a high flat rate but later restructured its pricing to lower the barrier to entry considerably.
The specific figures for both (ARR, pricing) swing widely by period and source, so here we note only the direction. The common thread is clear: they lead with a narrative of replacing or assisting "one employee," and they try to attach billing to output and usage rather than seats. This narrative sells well because it changes the buyer's math. The conversation shifts from "the cost of buying yet another tool" to "how much it costs to get one person's worth of work done."
Collaboration tools like Asana, ClickUp, and monday.com are each bolting on AI agents. See them only as "simple assistants that help with tasks inside the tool" and you'll miss their actual scope. ClickUp describes agents that take on the PM role, send messages and emails directly, and carry out work autonomously; monday's Project Manager agent organizes goals, scope, and milestones and drafts plan documents. On the axis of "performing a PM role and communicating like a person," their capabilities overlap considerably with ONTLE.
Here's where lumping the price into "X dollars per user per month" creates a misunderstanding: the base subscription fee and the AI agent fee are separate. ClickUp's agents and monday's AI run on a model where you consume credits based on usage, separate from your seat subscription. Each plan includes a certain allotment of credits, and what happens once you use them up isn't a simple automatic surcharge. ClickUp restricts the AI features when the included credits run out; to keep using them you either purchase additional credits or turn on a separate overage setting. monday likewise limits usage when credits are depleted and points you toward increasing your credit allotment. Buying extra credits (or enabling an overage option) and being billed automatically for going over are two different things. So a single line reading "$X a month" doesn't capture the actual bill. The final cost varies a great deal depending on how hard the team runs the agents.
The message this structure sends is this: AI features are no longer "free extras that come with the subscription" — they're becoming a resource priced by how much you use. Whatever services you compare, if you don't separate the base fee from the usage fee, the numbers will lie to you. A comparison table that folds three services into a single unit price may look clean, but it's no help in predicting the real cost of adoption.
Rather than nailing down specific statistics, it's more accurate to read the flow. A few directions stand out.
The backdrop is cost structure. Traditional SaaS enjoyed high margins because once it was built, adding more users cost almost nothing extra. AI SaaS, by contrast, incurs a model-call cost every time a request comes in. Because cost piles up the more it's used, how you control usage and how you bill for it becomes a question of profitability itself. That's why it's hard to push through on a single flat rate — and why the industry keeps experimenting with hybrid, credit, and outcome models.
Lay out the landscape and ONTLE's place comes into view.
| Item | Big Tech dispatch | Existing PM tools | ONTLE |
|---|---|---|---|
| Primary target | Mid-sized and up (per announcements) | SMBs to mid-market, all industries | Small teams, at the project level |
| AI's role | Technical implementation and engineering | Task assistance + a PM-role agent | Dedicated project PM |
| Customer touchpoint | Engineers/teams embedded on site | Using the tool + agents sending messages and emails | A dedicated AI PM as the single point of contact |
| Starting unit | Organization-level contract | Seat/credit subscription | Per project (can start small) |
| Persona | Function/organization-centric | Tool-centric | An AI PM with a name and a title |
Big Tech's dispatch-style services go in at the organization level and take over engineering. Existing PM tools now bolt on agents that take the PM role too — communicating by message and email, even producing plan documents — so on the "PM role and human-like communication" axis, there's a clear overlap with ONTLE. The difference ONTLE is after lies in how it packages that capability. At the unit of a single project, an AI PM with a name and a title takes sole ownership of that project, standing directly in front of the customer to receive requests and keep things moving — not as one feature inside a tool, but as the project's point of contact itself.
This isn't a story of "picking up a market Big Tech won't touch" or "doing what no one else can." Agents that stand in for the PM role already exist in several forms. But the three types differ in who they target, in the unit they start from, and in how they package that capability. Big Tech aiming at mid-sized companies and up doesn't mean it excludes small teams, and ONTLE focusing on small teams doesn't mean it goes head-to-head with large organizations. ONTLE's choice is "small, at the project level, with a PM who has a name and a title standing as that project's single point of contact." Instead of slugging it out over big organizations' large contracts, it bets on lowering the barrier to starting and making the point of contact feel closer to a person. Finding the open space in a competitive landscape doesn't start with doing what no one else can — it starts with solving the same problem from a different angle.
No. The announcements from Anthropic (May 4), OpenAI (May 11), and Microsoft (July 2) are spread across two months, and the cited investment figures vary by source. Above all, announced or raised amounts are plans, not completed deployments. Adding several announced figures together and squeezing them into one month is inaccurate.
Going strictly by the official announcements, you can't say that. Anthropic stated that its new company targets mid-sized companies. It's true they have an edge in large contracts, but no official material backs the claim that "small and mid-sized businesses aren't an option."
It's hard to answer in one line, because the base subscription fee and the AI agent fee are separate. ClickUp, monday, and others consume credits by usage, separate from the seat subscription. When the included credits run out, the AI features are limited rather than billed automatically — to keep going you purchase additional credits or enable a separate overage setting. The actual bill depends on how much you use the agents.
Big Tech stands in for engineering at the organization level. Existing PM tools also bolt on agents that take the PM role, communicating by message and email and producing plan documents, so on the "PM role and communication" axis they overlap with ONTLE. The difference is in how it's packaged. ONTLE works at the unit of a single project, where an AI PM with a name and a title becomes the sole point of contact owning that project. What differs is the starting unit, and how the capability is packaged and placed in front of the customer.