Skip to content
MakerGain
← Back to cases

After four hours a day watching dashboards, he gave AI a rollback button

Edited by MakerGainPublished: Article updated:

Plurio founder Seva Ustinov has spent more than twenty years in performance marketing. The product takes over button-pressing after marketers read reports, including restarting paused campaigns subsequently shown effective by attribution. Rather than immediately letting AI alter accounts, he made rules writable in natural language, previewable before execution, and selectable for human approval or automation, with logs and rollback. Early on, he used his own agent to analyse underperforming creatives and produce kill rules, planning to return to Cursor if it crashed. He did not switch back.

Performance marketers spend about 4 hours a day doing the same thing: watching dashboards, comparing campaigns and deciding which budget to increase and which campaign to pause. That was Plurio founder Seva Ustinov's description in a May 29, 2026 Starter Story interview. The people responsible for user growth also manage ad accounts, business analytics and traffic, while leading media buyers and agencies.[1]

Ustinov had worked in performance marketing for more than 20 years, led a team that once exceeded 100 people, and built an analytics product. The conventional next step would have been fuller, faster reports. His conclusion was the opposite: what held clients back was how quickly marketers made decisions after processing data, rather than the volume of data itself.[1]

Plurio was built around that conclusion. Its website says the AI agent compresses a typical 4-hour analysis into 2 minutes, targeting teams with annual paid-ad budgets above $5M across Meta, Google and TikTok. That is official marketing language, not a test result.[2]

It takes over the button-pressing stage

One listed action is easy to overlook: restarting a paused campaign that subsequent attribution confirms was effective.Source[2] It addresses a difficulty in long-funnel businesses. Ustinov says the niche's main problems are long funnels, delayed conversions, off-site revenue and long LTV, all requiring prediction and judgment that older software could not handle when making decisions.Source[1] A conversion may take weeks to appear; the ad paused on the spot may have been the wrong one to stop.

Before a machine directly changes money in an account, though, the product must answer what happens if it gets it wrong. The Rules module splits this into three steps: write a rule in natural language, preview the affected campaigns, ad sets and ads, then choose preview only, human approval or automatic execution. The company calls the last mode Full Automation: validated rules can run 24/7 with logs. Every action is recorded and changes can be rolled back. It says instructions are converted into deterministic code to avoid model hallucinations.[3]

Image: official product page[2], obtained on 2026-11-18, retaining the original interface; a website example, not a test result.

The rule examples are concrete. Pause a new Meta creative if CPA exceeds target after its first 200 impressions. Increase an ad set's budget by 20% when predicted ROAS, including delayed conversions, exceeds 3.0. Shut down a creative whose ROAS remains below break-even for two hours and replace it with the next-best performer.[3]

The trade-off deserves attention. Another ad-management tool, AdsMAA, also offers one-click pausing, proportional budget changes and targeting changes. Actions are preset by issue type and severity; custom action mapping is still planned. A pause can be undone by reactivating, while a budget change can be applied again or edited in Campaigns.Source[4] Plurio lets customers write rules in natural language or have AI propose best-practice rules for the business. In the Finom case, the customer team defines and owns CAC thresholds.

First, use your own product for a real job

Ustinov did not jump straight to having AI press buttons. In 2023, he and Kirill Kasimskiy founded Elly Analytics, a cross-platform marketing analytics business that supported more than $100M in annual ad spending and was later incorporated into Plurio. The company was renamed Plurio only after its March 2026 funding, so 2023 was not the beginning of the Plurio product.Source[5] He says that new models capable of planning, retaining context, using tools and evaluating themselves appeared in late 2024, and that he sat down to explore the work deeply on January 1, 2025.[1]

He described a recurring pattern inside client teams: “Most days were 10% thinking and creating, 90% clicking between platforms.”Source[5]

Asked about memorable moments, he gave two. The first came when the product was early and unstable. He still used Cursor for most tasks, but once used his own agent to solve a real problem: generating analysis and kill rules for underperforming creatives. The plan was simple—try the agent, return to Cursor if it broke. “I never switched back. It completed the whole thing.” He said this was his first realisation that it was a product, not a prototype.[5]

In 2025 they assembled a dedicated AI team and launched pilots, with the first customers using the agent for real budget decisions. His second moment came after the funding announcement. The company ran webinars and live sessions attended by people with clear intent. He said the shift from pushing a story to seeing pull was perhaps the strongest feeling.[5]

Who pays

Advertiser teams pay. The website targets teams spending more than $5M annually on paid advertising. The Pulse 2.0 interview introduction describes managing ad spending starting at $500K a month; when discussing the addressable market, the founder uses US companies spending $500K or more monthly. Those definitions differ and should not be treated as one threshold.[2] · Source[5] The website's “$500M+ in annual customer ad spend managed” means customers' advertising expenditure, not Plurio revenue.

The only visible revenue amount is self-reported: the May 29, 2026 Starter Story page lists $150K revenue/mo in its data field. That is interview-page data, not audited financials. Customer count, ARPU, gross margin and acquisition costs are undisclosed, leaving profit unclear.Source[1] Plurio's own subscription prices also do not appear on the public pages.

For outcomes, the Finom customer story provides one set covering November 2025 to March 2026: new customers +38%, CAC −33%, total spending −8%. Two sections on the same page offer other sets: core channels show new customers +35%, CAC −17%, spending +12%; high-frequency experiments show new customers +100%, CAC −57%, spending −13%. Their scopes and baselines differ and cannot be merged into one result.[6]

The TripleTen case appears in the Pulse 2.0 interview. Its US performance team spent more than $1.5M monthly on Meta, Google, YouTube and TikTok. Campaign analysis fell from 1–1.5 hours to 10–15 minutes, saving 20 hours a month, with full adoption taking two weeks. By March 2026, 11 automated rules were running across Search, YouTube, Facebook and creative-fatigue detection. The agent could propose and directly execute budget increases, reductions and creative pauses in ad accounts.Source[5] The founder also said that, at TripleTen's rates, analysis time saved alone covered the product fee. Finom's numbers come from Plurio's customer-story page; TripleTen's come from the founder's Pulse 2.0 account. Neither has a third-party audit.

Ustinov also says Plurio operates AI-first internally. Its 30-person team shares an AI workspace, with AI subscriptions and tokens costing about 4% of payroll. He expects that to reach 10% within a year and claims a 2- to 3-fold productivity improvement.[5]

Another of his judgments is that product-market fit is no longer an arrival point: it lasts about three months. Because LLMs, Claude Code and competitors have access to the same tools, the company must reinvent, build and release new value every three months.[5]

This company raised $3.5M in March 2026, bringing total funding to $4.5M, and self-reports $150K in monthly revenue. The moat is not the model itself, but how quickly it becomes a button customers dare to press. Plurio makes that button adjustable: preview first, then human approval, then 24/7 execution of validated rules, with logs and rollback at every step. It sells a willingness to let go gradually, rather than surrendering the entire account at once.

Continue reading

AI productivity
Over 180,000 views, 1 February sale: AI Renamer leaves batch-renaming confirmation to people

Ozgur Ozer's AI Renamer gives content-based names to opaque filenames such as DSC_0412.JPG and scan_20250312.pdf. For manual batches, people inspect the suggested names before pressing Rename All; nothing changes on disk beforehand. Ozer had always given away his small apps and first used Polar to solve payment collection because Turkey had no Stripe support. His first post exceeded 180,000 views but produced 1 sale that month; sales began coming after local support and a Product Hunt launch.

AI productivity
After closing a product nobody wanted, they broke entrepreneurship into “the next step”

David Heikka and his brother Felix launched their first AI form product and found zero demand. They tried ads, landing-page changes, renaming, and added features, then closed it after six months. They went on to build Buildpad, breaking business creation into stages that indicate what to validate next. Its early and current billing mechanisms differ, and the product was renamed aicofounder before changing back to Buildpad.

AI productivity
It writes your emails, but stops at LinkedIn

According to its website’s workflow, DeepReachAI finds people, researches recent developments, writes messages, classifies replies, and books meetings for cold outreach, while excluding automatic LinkedIn sending, WhatsApp scraping, and calls without consent. In an IndieHackers post, founder Aryan Agrawal said he had expected personalization to be the hook, but the compliance boundary—no automatic LinkedIn sending and no WhatsApp scraping—was what made people connect accounts. The homepage says Orio can send a first email through a connected account, without identifying the account provider.

AI productivity
An assistant cost thousands of US dollars a month. He put AI to work and reached $10k+ MRR

Because an assistant would cost thousands of US dollars a month, Elie Steinbock built an AI email assistant instead. It connects to an inbox, sorts messages, and drafts replies using email history, calendars, and customer records; a person still presses Send. On Product Hunt launch day, he emailed 20,000 newsletter subscribers and posted a video on his YouTube channel with 2,000 subscribers. He estimated that these extra efforts may have added fewer than 50 votes—a judgment, not a measured result.