Boaz Hecht Is Drawing the Map AI Agents Still Need

A refund request looks simple from the outside. Inside a large company, it can be a tiny expedition. A support agent opens the ticket, finds the order, checks a payment system, consults a policy, makes a judgment and writes back. Each stop may live in a different application. Each company may have customized those applications. The official instructions may describe a route that the best agents stopped taking months ago.

Boaz Hecht has built his latest company around that distance between the neat diagram and the actual journey. At 8Flow, the co-founder and CEO talks about “micro workflows” - the granular sequences of clicks, lookups and decisions that accumulate into enterprise work. His pitch is that AI agents will not become useful merely because they can press the buttons. They need the local map: which button, in which system, after which check, under which company rule.

Hecht reaches for a consumer metaphor to explain the missing layer. 8Flow, he says, is trying to build something like Waze for enterprises. The roads are the systems a company already uses. The traffic is the live behavior of employees. The destination is a resolved case. And the route changes from one organization to the next.

We basically decided to build something along the lines of Waze for enterprises.Boaz Hecht, 2025

A problem that kept changing clothes

The mapmaker framing is new. Hecht’s fixation on awkward enterprise work is not. His public career begins around 2000, when he co-founded ClubSMS while studying computer science and management at King’s College London. Consulting and business-development roles followed, including work at Deloitte, SRL Group and Cornerstone Innovations. By 2012, he was back in the founder’s chair with Itay Braun at SkyGiraffe.

SkyGiraffe attacked a problem that sounds quaint only until you remember how slowly corporate software moves: workers expected polished mobile apps in their personal lives, while enterprise systems still pinned them to a desktop. The company connected back-end business applications to native mobile experiences. Its stated recipe was almost aggressively plain: get the data, show it and let people update it, while keeping everything secure enough for the CIO.

The name came first. Late on May 24, 2012, Hecht and Braun decided they liked giraffes, liked the sky and wanted something easy to brand. Hecht later admitted the story was anticlimactic. The work was not. In his account, the team was intense, customer-minded and “kind of maniacal” about the product and service. Selling a platform meant running alongside much larger companies, or, in the warning he remembered receiving, running with elephants.

In October 2017, one of those elephants picked SkyGiraffe up. ServiceNow announced an agreement to acquire the company and embed its technology into the Now Platform. The fit was practical: ServiceNow wanted native mobile experiences, while SkyGiraffe had spent five years turning complicated enterprise systems into usable ones. Hecht wrote that there was nearly complete overlap between the startup’s customers and ServiceNow’s.

He stayed. Over roughly five years, Hecht moved from acquired founder to VP of Platform, leading areas that included the core platform, mobile, web, conversational products and employee-facing experiences. It gave him a view of both sides of enterprise software: the small team building one sharp abstraction and the large platform responsible for thousands of customers, customizations and edge cases.

The second founding team

By late 2021, Hecht wanted to build again. He did not assemble an unfamiliar cast around a fashionable category. He returned to Josh Russ and Yev Goldin, colleagues from SkyGiraffe who had continued through the ServiceNow acquisition. They started 8Flow in 2022. By Hecht’s count in 2025, the founding team had been working together for 11 years.

The company initially focused on customer support. This was a market the founders knew, and its daily work made the problem visible. A support agent might switch among Zendesk, ServiceNow, Salesforce Service Cloud, spreadsheets and internal tools just to complete one routine case. 8Flow’s early Chrome extension learned repeated steps and offered one-click actions, saving people from copy-and-paste loops while collecting the feedback needed to understand what came next.

Hecht described the learning process with Lego bricks. First observe which pieces need to exist. Then learn how people assemble them. Eventually, use the resulting structure to reproduce a workflow. It was a bottom-up answer to the brittleness of traditional automation, which often asks consultants or employees to document every rule before anything can run.

The approach also reveals errors that a ticket log can miss. Hecht offers the example of an outdated knowledge-base article. Veteran agents may ignore it because they know the procedure. New hires follow it and become slower or wrong. A dashboard can show that one group takes longer. A workflow map can show exactly where the paths diverge.

Waiting for the action layer

8Flow emerged from stealth in June 2023 with a $6.6 million seed round led by Caffeinated Capital. Its near-term product already had a reason to exist: fewer tab switches, fewer repetitive actions and better visibility into which tools teams actually used. The long-term bet was that these observations could train models rich enough to generate workflows tailored to each user and company.

The timing required patience. For its first two years, Hecht said, the team concentrated on the pipeline that collected, structured and analyzed workflow data. It was preparing the map while the vehicles were still unreliable. As browser automation and AI agents improved, that stored context became more valuable. OpenAI, AWS, Anthropic and others could build ways for agents to act; 8Flow would tell those agents what to do and why.

In April 2025, the company announced a $10 million Series A. Caffeinated Capital led again, joined by strategic investors ServiceNow Ventures and Okta Ventures, plus HNVR, Liquid 2 Ventures, FirsthandVC and angel investors. Hecht said 8Flow still had most of its prior round in the bank. The raise was less a rescue than a decision to accelerate as the surrounding technology caught up.

We’re building the intelligence behind it that tells the AI what to do and why.Boaz Hecht, 2025

By then, 8Flow had tracked more than 600 million workflow events. Hecht’s ambition had expanded from saving a support agent a few clicks to supplying the context layer for enterprise agents. A password reset, in his telling, might touch ServiceNow, Workday and Okta, with a compliance check between them. A capable agent can operate each tool. A useful agent must understand why the sequence exists and which steps can happen in parallel.

Odd names, serious continuity

For someone building maps, Hecht is refreshingly relaxed about origin mythology. The explanation for 8Flow echoes SkyGiraffe’s. The founders liked the number eight. It looked a little like infinity. Infinite workflows made for a plausible story, but Hecht called that story partly reverse-engineered. “We just liked the name,” he said, adding that after SkyGiraffe, the next company had to remain slightly odd.

The serious continuity is found elsewhere: in the colleagues who followed one another across three companies, in the investors who returned for another round, and in a problem that became more interesting as the technology around it changed. SkyGiraffe helped people reach enterprise systems from a phone. ServiceNow showed the team what those systems look like at platform scale. 8Flow is trying to capture the routes people improvise through them.

There is also restraint in the division of labor. Hecht does not frame 8Flow as the company that must build every agent, model or browser action. The larger providers can supply the motion. His company wants to supply judgment-shaped context: the destination, the allowed roads, the alternate routes and the local rules that never made it into the manual.

That leaves 8Flow with a difficult, unglamorous assignment. It has to watch without adding friction, separate a durable pattern from noise, and convert human practice into a structure machines can use. It must respect the nuance that makes each enterprise different while finding enough repetition to make a product. The work sits below the agent demo, where the hard part is not making the cursor move. It is making the movement belong.

Hecht’s career makes the wager feel less like a sudden AI pivot than the latest version of a long observation. Corporate work spills across interfaces. People patch the gaps with memory and habit. Every new computing layer promises to simplify the mess, but first it has to understand it. For 8Flow, the route to automation begins one level earlier - with the patient work of drawing the road.