Runable, a fifteen-person startup founded in Bengaluru in 2025, has raised $21 million in a Series A that values it at $65 million after the money. That is a small round by 2026 standards and a very large one for a company that only started charging customers in March. The interesting part is not the cheque. It is the argument the cheque is buying.
The argument is this: building a website, an app or a deck with an AI agent has already become a commodity, and the money is in what happens next. TechCrunch reported the round on 26 August 2026, and the framing its co-founder gave is unusually blunt for a founder in a hot category. Businesses do not want a coding tool. They want customers.
This article breaks down what Runable actually announced, the numbers behind it, the economics the company openly admits are not working yet, who it is fighting, and — the part most coverage skips — what any of it should change for a small business that is buying artificial intelligence rather than building it. There are figures here that flatter the company and figures that do not, and both are worth reading before deciding whether a general agent belongs anywhere near your marketing budget.
Table of contents
- What Runable Announced on 26 August
- Who Runable Is and Where It Came From
- The Bet: Runable Thinks Building Is the Easy Half
- The Numbers Runable Put on the Table
- What Runable’s Agent Actually Does
- Runable’s Pricing and the Stack It Wants to Replace
- The Problem Runable Has Not Solved: Negative Gross Margins
- Runable’s Competition Is Everyone at Once
- Why Growing a Business Is Harder Than Building One
- Where Runable Sells: The US, the UK, Japan and Brazil
- The Bengaluru Angle: Runable in India’s AI Cohort
- What Runable Means If You Are Buying AI, Not Building It
- Risks and Open Questions for Runable
- What Runable Has to Prove in the Next Twelve Months
- Frequently Asked Questions About Runable
- References
What Runable Announced on 26 August
The announcement is compact, and the details matter more than the headline number.
The round in one line
Runable raised $21 million in Series A funding. The round was all-equity and entirely primary, meaning every dollar goes onto the company balance sheet rather than into an existing shareholder’s pocket. Post-investment, the company is valued at $65 million.
Who led it
The round was co-led by Susquehanna Venture Capital and Nexus Venture Partners. Existing investors Together Fund and Array VC also participated. Susquehanna is the venture arm of a large American trading firm; Nexus is one of the longer-established India-US crossover funds. The pairing tells you the company is being underwritten as a global consumer-scale software bet, not an India-domestic one.
What kind of money it is
A primary, all-equity round at this stage usually means the founders are not taking money off the table and the investors are betting on a step change rather than a steady climb. At $65 million post-money, the $21 million represents roughly 32 per cent of the company. That is a heavy dilution for a Series A, and it says something about how much capital the business model needs.
The company at a glance
| Detail | Figure |
|---|---|
| Round size | $21 million Series A, all-equity, all primary |
| Post-money valuation | $65 million |
| Co-leads | Susquehanna Venture Capital, Nexus Venture Partners |
| Existing investors participating | Together Fund, Array VC |
| Founded | 2025, Bengaluru, India |
| Headcount | 15 |
| Registered users | Approximately 1.7 million |
| Annualised revenue run rate | $2 million |
| Gross margin | Negative |
Who Runable Is and Where It Came From
A company barely a year old raising at $65 million invites a fair question about substance.
A 2025 company with a 2026 balance sheet
Runable was founded in 2025. Payments launched in March 2026. That means the entire commercial history of this business is measured in months, and the Series A was priced on a revenue line that has existed for less than half a year. Investors were buying a growth curve, not a track record.
The founders
The company is led by co-founder and chief executive Umesh Kumar, alongside co-founder Saksham Sarda. Kumar has been the public voice of the round and the one putting numbers on the record, including the ones that do not look good.
Fifteen people, 1.7 million users
The headcount is the number worth staring at. Fifteen employees serving roughly 1.7 million registered users is a ratio that would have been science fiction in 2020. It is also a direct consequence of the product being an agent: the software does the work that a services company would otherwise hire for, and the marginal cost of a new user is inference, not payroll.
Where the users actually are
Runable’s largest markets are the United States, the United Kingdom and Japan, with Brazil also material. Kumar expects Japan to sit alongside the US at the top before long. For a Bengaluru company with no obvious distribution advantage in any of those markets, that geographic spread is either evidence of genuine product pull or evidence of very efficient paid acquisition. It is not yet possible to tell which from the outside.
The Bet: Runable Thinks Building Is the Easy Half
Every AI company has a thesis. This one is unusually falsifiable.
The build layer is commoditising
Natural-language app building went from novelty to crowded category in under two years. Cursor, Lovable and Replit all sell it. Anthropic and OpenAI both ship agents that do a version of it. The marginal value of being the fifteenth product that turns a prompt into a landing page is close to zero, and Runable appears to have concluded that before its investors did.
What “grow” means in practice
So Runable is extending the agent past the build step. The stated targets are running advertising campaigns, managing social media, handling search optimisation, and optimising how a business appears inside AI chatbot answers. That last one is a genuinely new surface. When a buyer asks an assistant for a recommendation instead of typing a query into a search box, being absent from that answer is the modern equivalent of being on page four.
Kumar’s framing
The chief executive put it plainly to TechCrunch: “In the end, a business doesn’t require Codex or Claude Code or anything. They require real outcomes.” Whatever you think of the product, that sentence is a fair description of what most small businesses feel when they open a developer tool.
The pricing logic behind it
The second quote is the commercial one. “If I am paying an agency $10,000 to run my Google Ads, can someone come in and do it for me for a lower price? That’s where Runable comes in.” The comparison is not to another software subscription. It is to a services invoice, and services invoices have far more room in them.
Build versus grow, side by side
| Dimension | Build side (today) | Grow side (the bet) |
|---|---|---|
| Typical output | Website, app, deck, report, sheet | Ad campaigns, social posts, search and chatbot visibility |
| Who it replaces | A freelancer or a template tool | A marketing agency retainer |
| Budget it competes with | Tens of dollars per month | Thousands of dollars per month |
| How success is judged | Does the artefact exist and work | Did revenue move |
| Failure is visible | Immediately | Weeks later, after money is spent |
| Competitive density | Very high | Lower, for now |
The Numbers Runable Put on the Table
Kumar disclosed more operating detail than most founders at this stage, which is worth crediting even where the numbers are unflattering.
Two million dollars in three weeks
Runable went from zero to a $2 million annualised revenue run rate within three weeks of switching on payments in March. Read that carefully: it is a run rate, not $2 million collected. It annualises a short window of early-adopter demand, and early-adopter demand is the least durable kind. It is still a fast start.
One trillion tokens in ninety days
Users consumed more than one trillion tokens in the last ninety days. That works out at roughly 11 billion tokens a day, or about 588,000 tokens per registered user across the quarter — a little over 6,500 tokens per user per day if you spread it evenly, which of course it is not. It is a real usage number, not a signup number, and those are rarer.
The paying share
Between 60 and 70 per cent of that token consumption comes from paying customers. That is a healthy ratio in a freemium product and suggests the free tier is not being strip-mined at the scale the raw user count might imply. It also means roughly a third of a very large inference bill is being spent on people who pay nothing.
What the valuation implies
At $65 million post-money against a $2 million run rate, Runable is priced at roughly 32.5 times revenue. For context on the same arithmetic, Replit has been reported at $265 million of annualised revenue against a $9 billion valuation, about 34 times; Lovable at roughly $500 million against a reported $13.2 billion, about 26 times. On that single measure, this round is priced squarely in line with the category rather than at a premium.
What Runable's Agent Actually Does
Strip the funding story away and there is a product underneath, which is more than can be said for some rounds.
Natural language in, artefacts out
The agent takes a prompt and produces websites, apps, slide decks, documents, images, carousels, reports, spreadsheets, audio and video. That range is deliberately wide. Runable is not positioning as a coding tool that happens to make slides; it is positioning as the single place a small business goes instead of opening five subscriptions.
The natural language processing layer is doing the same job a project brief would: turning a vague sentence into a specification the system can act on. Where that fails, it fails quietly, and the user gets a confident-looking artefact built on the wrong assumption.
The infrastructure it hides
The part that matters commercially is what the user does not see. Runable handles deployment and analytics itself, so a non-technical owner never meets a hosting dashboard. Every one of those artefacts is produced by a large language model the company does not own, which is the hinge the entire margin question turns on.
The benchmark claim
The company advertises a 92.1 per cent score on GAIA, an academic benchmark for general assistants that tests real-world reasoning and tool use rather than trivia recall. Benchmark scores are marketing, and GAIA in particular has been climbed hard by every serious agent vendor. Treat it as evidence the product is competent, not as evidence it is best.
Where it runs
There are clients for iPhone, Android, Mac and desktop. For the small-business audience, mobile is not a nice-to-have — a lot of the target market runs its marketing from a phone between other jobs, and a browser-only agent quietly excludes them.
Runable's Pricing and the Stack It Wants to Replace
The price is the strategy, and it is set low enough to be a statement.
Fifteen dollars a month
Runable lists a subscription at $15 per month billed annually, or $20 per month on standard billing. That is consumer pricing. It is below what most agencies charge for an hour, and it is roughly one two-hundredth of the $10,000 Google Ads retainer Kumar used as his reference point.
The claimed saving
The company’s own comparison claims a user saves $139 a month — $1,668 a year — versus subscribing separately to the tools it replaces, naming assistants and design tools among them. Take the vendor’s arithmetic at face value and the replaced stack costs about $154 a month, of which Runable proposes to charge under ten per cent.
Why price is the whole strategy for small businesses
Small firms do not buy software on capability, they buy it on the size of the cheque and the number of logins. A single $15 line item that removes four subscriptions and a freelancer invoice is an easy internal argument. A $200 platform that does the same job better is a project, and projects do not get approved.
Where that price becomes a problem
Consumer pricing against inference costs is exactly how you end up with the margin position described in the next section. Runable is buying market share with someone else’s compute, and the round is what funds the gap.
The Problem Runable Has Not Solved: Negative Gross Margins
This is the number the company volunteered, and it is the most important one in the story.
Subsidised inference
Runable’s gross margins are currently negative, partly because it subsidises AI usage. In plain terms, for at least some customers the cost of serving them exceeds what they pay. That is not a rounding error at the gross line — it is the line before you have paid a single salary.
Why the trillion-token figure cuts both ways
A trillion tokens in ninety days is impressive engagement and an enormous bill. The same number that proves the product is used is the number that proves it is expensive, and roughly a third of it comes from users generating no revenue at all. Growth in usage, at this price point, currently makes the loss bigger.
The plan
The company’s answer has two parts: build its own models to displace some third-party inference, and wait for inference prices to keep falling. Both are reasonable. Neither is under Runable’s control on any particular schedule, and a fifteen-person team training competitive models is an ambitious sentence.
What has to be true
For this to resolve, one of three things must happen: inference costs fall far enough to flip the unit economics, the paid mix improves enough to cover the free tier, or prices rise. The third option is the one that quietly kills consumer-priced products, because a $15 subscription cannot become a $60 subscription without losing the customers who chose it for the price.
Runable's Competition Is Everyone at Once
The competitive set here is not a list. It is three separate industries arriving at the same customer.
The model labs
Anthropic and OpenAI both ship agentic products that overlap directly with the build side. They have the models, the distribution and no need to earn a margin on inference they already own. Any feature Runable ships that proves popular is a weekend for them.
The coding platforms
Cursor, Lovable and Replit own the developer-adjacent end. Lovable reportedly passed $500 million of annualised revenue by June 2026, up from around $200 million in November 2025, and Replit has been reported at $265 million. Runable is not competing with these on scale; it is competing on audience, targeting the owner who would never open a code editor.
The general agents
Manus and Genspark occupy the closest ground — general-purpose agents that produce mixed artefacts for non-developers. This is where the fight actually is, and it is the segment with the least product differentiation and the most price pressure.
The incumbent nobody lists
The real competitor is the local agency charging a monthly retainer, and the freelancer charging by the hour. Those relationships lose on price and win on accountability, and accountability is a hard thing for a chatbot to sell.
How the field lines up
| Player | Core audience | Overlap with Runable | Structural advantage |
|---|---|---|---|
| Anthropic, OpenAI | Everyone | Build side, increasingly the grow side | Owns the models and the inference cost base |
| Cursor | Professional developers | Low | Deep workflow lock-in with engineers |
| Lovable, Replit | Prosumer app builders | High on build, low on grow | Scale, brand and very large revenue base |
| Manus, Genspark | Non-technical knowledge workers | Direct | Same thesis, similar stage |
| Local marketing agencies | Small and mid-sized businesses | Direct on the grow side | Accountability, relationships, local knowledge |
Why Growing a Business Is Harder Than Building One
The thesis is right about the opportunity and probably underestimates the difficulty.
Building has a visible finish line
When an agent builds a landing page, you can look at it. It is either there or it is not, it either loads or it does not, and a bad result costs you an afternoon. The feedback loop is immediate and the downside is bounded.
Growth has no acceptance test
When an agent runs a campaign, the output is a spend and a set of numbers that may or may not be attributable to it. There is no moment where the work is obviously finished and obviously correct. Judging whether it worked requires a baseline, a time window and a tolerance for noise, none of which a small business typically has.
Spending real money on someone else’s behalf
There is a categorical difference between an agent that writes copy and an agent that has authority over an advertising budget. The first can waste your time; the second can waste your quarter. Every serious deployment of autonomous spend needs limits, approvals and an audit trail, and building those well is a product in itself.
The trust gap
An agency that burns your budget answers the phone. That is not a small part of what the retainer buys. Until an agent can carry equivalent accountability — a named owner, a documented decision trail, a route to recovery — the price advantage will not be enough for the customers with the most to lose.
Where Runable Sells: The US, the UK, Japan and Brazil
The geography is the most quietly interesting fact in the whole announcement.
An Indian company with a Western revenue base
Runable is built in Bengaluru and monetised in the United States, the United Kingdom and Japan. That inverts the usual pattern for an Indian software startup, where the domestic market is the proving ground and the West comes later. Here, the product went straight to the highest-willingness-to-pay markets.
Why Japan is the surprise
Japan rarely appears in an early-stage AI startup’s top three. Kumar expects it to sit level with the United States before long. Japanese small businesses face an acute labour shortage and a long-standing gap in digital marketing capacity, so an agent that does the work of a small agency has an unusually clean value proposition there — if the language quality holds.
What that implies about the product
A user base spread across four language markets with fifteen employees means almost no human-in-the-loop support and almost no localisation team. That is only possible because the agent itself does the localising. It is efficient, and it is fragile: quality problems in a market you do not staff are invisible until churn shows up.
Brazil, and the price-sensitive tail
Brazil rounds out the picture as a large, price-sensitive market where $15 a month buys meaningfully more than it does in London. Volume from markets like that flatters user counts and does comparatively little for revenue, which is worth remembering when reading the 1.7 million figure.
The Bengaluru Angle: Runable in India's AI Cohort
The round also says something about where Indian AI companies are being priced in 2026.
A crowded, fast-moving cohort
India now has a large cohort of AI-native startups shipping globally from day one, and Bengaluru is the centre of it. What has changed is not the ambition but the go-to-market: products are launched worldwide, priced in dollars, and sold self-serve, so the domestic market is no longer a prerequisite.
What a $65 million valuation signals
Sixty-five million dollars post-money is not a headline valuation in 2026. Against a $2 million run rate it is a considered price, not an exuberant one, and the fact that the round is entirely primary with a third of the company sold suggests investors priced the margin risk rather than ignoring it.
The capital-efficiency argument
Fifteen people, four significant markets and a functioning revenue line is the kind of ratio that makes a modest round go a long way. Whether that efficiency survives the move into advertising operations — a domain where the failure modes cost customers money — is the open question.
What Runable Means If You Are Buying AI, Not Building It
Most readers of this article will never invest in a company like this. They will, however, be sold something very like it within the year.
The honest read for a small business
An agent priced at $15 a month is worth trying, because the cost of trying is a rounding error and the cost of being the last business without one is not. What it is not is a replacement for owning your own strategy. Tools that generate marketing output cheaply make the average output worse across the whole market, which raises rather than lowers the value of a considered plan.
Where an agent earns its money
The reliable wins are the repetitive, low-judgement, high-volume tasks: drafting variants, resizing assets, producing first-pass reports, keeping listings consistent, filling a content calendar. That is the same territory as any well-scoped business process automation project, and the discipline that makes those succeed applies here too — define the task, define what good looks like, and measure.
Where it will disappoint you
It will disappoint you on anything requiring context it does not have: your margins, your worst customers, the deal you cannot afford to lose, the regulatory line you must not cross. It will also disappoint on visibility inside AI answers, which is a moving target that rewards structured, authoritative source material rather than volume. If that surface matters to you, treat it as its own discipline — the same one covered by AEO services and GEO services — rather than a checkbox in an agent.
The sensible middle path
Run the cheap agent for volume work, keep human judgement on anything that spends money or carries risk, and hold the strategy yourself. If you want the same leverage applied to defined roles rather than one-off tasks, that is the model behind AI employees and autonomous AI agents, where scope, guardrails and hand-off points are designed up front instead of discovered later.
Keeping track of the field
The category is moving weekly and most of the noise is funding announcements rather than capability changes. Our AI models and tools hub tracks the releases that actually change what a business can do, which is a much shorter list than the news cycle suggests.
Risks and Open Questions for Runable
Four things could go wrong here, and none of them are exotic.
Concentration risk on model providers
The company’s cost base sits with third-party model providers who are also, in Anthropic and OpenAI, its competitors. A pricing change, a rate limit or a terms-of-service revision upstream lands directly on the income statement. Building in-house models is the stated mitigation, and it is a long road for a small team.
Churn is invisible from the outside
A run rate established in three weeks tells you nothing about retention. The metric that decides whether this business exists in 2028 is what proportion of March’s paying customers were still paying in August, and that number has not been disclosed. In a consumer-priced product sold to small businesses, monthly churn in the high single digits is normal and quietly fatal.
The category has a high failure rate
Gartner has forecast that more than 40 per cent of agentic AI projects will be cancelled by the end of 2027, citing escalating costs, unclear business value and inadequate risk controls. That is a statement about buyers, not vendors, and it means the demand curve Runable is extrapolating may not be as smooth as the first three weeks suggested.
The opportunity is real even so
The same analysts expect 40 per cent of enterprise applications to include task-specific agents during 2026, up from under 5 per cent the year before, and have put $234 billion of enterprise application software spend at risk from agentic AI. Both things are true at once: the shift is enormous and most individual attempts at it will fail.
What Runable Has to Prove in the Next Twelve Months
The round buys time. Here is what the time has to produce.
The margin milestone
The single test is whether gross margin crosses zero without a price rise. Everything else is secondary, because a business that loses money on every customer does not become healthier by adding customers. Watch for any announcement about in-house models, and read it as a cost story rather than a capability story.
The retention milestone
The second test is a disclosed retention or net revenue figure. A company confident in retention publishes it. Continued silence past the first anniversary of payments launching would be the loudest signal available.
The differentiation milestone
The third test is whether the grow side ships as a genuine product rather than a set of prompts. Running advertising well requires budget controls, attribution, approval flows and a way to stop. If what arrives is a chat interface that writes ad copy, the thesis has not actually been implemented.
The scorecard
| Milestone | What good looks like | What failure looks like |
|---|---|---|
| Gross margin | Positive without raising the headline price | Usage caps quietly tightened, or price doubled |
| Retention | A published net revenue retention figure | Continued silence, user counts quoted instead |
| Grow-side product | Budget controls, approvals, attribution, a stop button | Ad copy generation described as campaign management |
| Own models | A measurable share of inference moved in-house | An announcement with no cost impact |
| Headcount | Scales far slower than revenue | Support hiring to patch product gaps |
Frequently Asked Questions About Runable
How much did Runable raise?
Runable raised $21 million in a Series A announced on 26 August 2026. The round was all-equity and entirely primary, meaning the capital went to the company rather than to selling shareholders.
Who invested in Runable?
The round was co-led by Susquehanna Venture Capital and Nexus Venture Partners. Existing backers Together Fund and Array VC also took part.
What is Runable valued at?
Sixty-five million dollars after the investment. That prices the company at roughly 32.5 times its stated $2 million annualised revenue run rate.
What does Runable actually do?
It is a general AI agent that turns natural-language prompts into websites, apps, presentations, documents, images, reports and spreadsheets, handling deployment and analytics behind the scenes. The company is now extending it to run advertising, manage social media, handle search optimisation and improve how a business appears in AI chatbot answers.
How much does Runable cost?
The listed subscription is $15 per month billed annually, or $20 per month on standard billing. The company claims that replaces about $154 a month of separate tool subscriptions.
Is Runable profitable?
No. Gross margins are currently negative, partly because the company subsidises AI usage. Management expects the position to improve as it builds its own models and as inference costs fall.
Who does Runable compete with?
On the build side, Anthropic, OpenAI, Cursor, Lovable and Replit. On general agents, Manus and Genspark. On the grow side, the real competitor is the marketing agency retainer it is priced against.
Should a small business use an agent like this?
At $15 a month it is cheap enough to test on low-risk, repetitive work. Keep human approval on anything that spends money, and do not confuse an agent that produces marketing assets with one that owns your marketing strategy.
References
Runable hits $21M to bet AI agents can go from building businesses to growing them
Together Fund portfolio: Runable
Nexus Venture Partners — Crunchbase company profile
Gartner Predicts 40% of Enterprise Apps Will Feature Task-Specific AI Agents by 2026
Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027
Gartner Says $234 Billion in Enterprise Application Software Spend Is at Risk from Agentic AI
AI Coding Startup Lovable In Talks To Raise Funding At A $12 Billion Valuation
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.