Why 99% Accurate Browser Agents Still Fail — Derek Meegan, Browserbase

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Why 99% Accurate Browser Agents Still Fail — Derek Meegan, Browserbase · Channel: AI Engineer · 16:46 · Watch on YouTube

Topics: AI agents, Task and workflow automation, AI evaluation and benchmarks, Software engineering with AI

Summary

Derek Meegan explains why browser-agent demos often fail to translate into reliable production systems: step-by-step costs and failure risks accumulate, while value is realized only when the entire task succeeds. He argues that teams should judge success by completed customer transactions, use retries, and verify results through concrete artifacts. After performance is established, cost and maintainability can be improved through engineering. His example shows how deterministic tools, reusable authentication, verification, and agent skills can limit ambiguity and improve reliability.

Key points

Browser agents interact with the web through representations such as HTML and accessibility trees, screenshots, or dynamic code execution.
Transactional browser workflows are difficult to scale because every step adds cost and risk, while the task’s value is realized only upon completion.
Even a 99% success rate per step can yield only about 36% success across a 100-step trajectory.
Measure success per transaction rather than per run, and allow retries when evaluating reliability.
Define success through a concrete artifact, such as a confirmation email, order ID, receipt, or queryable system record.
Assess performance first; once the agent can complete the task, treat cost and maintainability as engineering optimization problems.
Durability risks include anti-bot measures, changing tasks and websites, better methods becoming available, and models straying from the intended path.
Use deterministic tools to encapsulate complex operations and verify outcomes against a system of record.
Move stable workflows, such as authentication, outside the model’s responsibilities and reuse them where possible.
Provide skills that describe a task’s critical path, reducing ambiguity and the number of decisions the model must make.

Chapters

Start	Chapter	Summary
00:00	From browser-agent demos to production	Meegan introduces the gap between a successful demo and running browser agents reliably at scale, then outlines how agents interact with web pages.
03:08	Agent trajectories and the production challenge	He distinguishes purpose-built, intermediate, and just-in-time workflows, then explains why transactional tasks accumulate risk and cost before delivering value.
06:17	Defining value and measuring success	He frames agents as a business investment and emphasizes customer outcomes, concrete evidence of completion, and per-transaction success with retries.
09:26	Costs, maintenance, and durability risks	He breaks down operating and maintenance costs and describes risks including anti-bot measures, changing tasks, new methods, and model variability.
12:01	Building a more reliable browser-agent system	Using an insurance-portal example, he shows how deterministic tools, verification, reusable authentication, and skills can reduce ambiguity and improve performance.
16:11	Production and on-call responsibility	He concludes that reducing unnecessary model responsibilities makes the system more maintainable and performant, then welcomes the audience to production.

Quotes

Cost accumulation is continuous while value realization is terminal. — 04:41

If your browser trajectory spans 100 steps, the overall success rate at scale starts to look like 36%. — 05:13

profit equals revenue minus costs. — 06:17

permit retries and measure success against a per transaction basis. — 09:26

Now you’re on call. — 16:11

Concepts and entities

Derek Meegan, Browserbase, browser agent, DOM, accessibility tree, Computer Use, CDP, OCR, skill, browser automation

Full transcript (original language)

00:00 All >> [music] >> right. Deploying browser agents. Uh running a browser agent demo once is easy. Uh it’s on one site. It’s one run on a good day with you watching and your fingers crossed. Uh, but we know

00:32 production is vastly different. Thousands of runs unattended while the site can change or your model can just have an off day. So that’s why we’re here. What separates a demo from production? My name is Derek Megan. I’m a software engineer at Browserbase. And today we’ll break up this talk into four parts. First, we’ll bring order to chaos, and we’ll understand concretely how agents interact with the web. Second, we’ll

01:03 define the impossible and explore why running them at scale is so hard. Third, we’ll take a step back. We’ll think about what’s important and the value proposition that agents propose. And then last, which is hopefully the reason why we’re all here, we’ll generate some shareholder value. So first how agents interact with the web. I like to think about the output of a model as a probability distribution. You receive input tokens. They represent the

01:35 page state, the overarching goal or the steps that have already been taken. Then the model explores a distribution of possible actions proposing the most likely one. Then that action is transposed into a structured tool call and deterministically executed to interact with the browser. But the interface of the browser is complicated. You have several distinct layers. You have requests flowing in and out of the browser. You have the document object model or the HTML

02:05 representation of the page. You can screenshot the browser. the accessibility tree which is a semantic textual representation of the page and the browser state storage cookies console URL tabs etc. Underneath all of this, you have a dynamic code execution runtime where you can interact with the browser by writing arbitrary do JavaScript or CDP. How do you wire these together? Typically in industry, there’s three

02:36 main strategies. Number one is creating a textual representation of the browser. This typically involves taking the HTML state of the page in the accessibility tree and creating a hybrid representation. Number two is computer use where you represent the browser in a series of screenshots of the page. And finally, and becoming increasingly popular, is deharing your agent, removing tailor made tools in favor of dynamic execution environments where the agent can write arbitrary code against the browser.

03:08 In practice, most production systems lean on the first, the second, or a combination of the first and the second two strategies. Now let’s think about the type of agentic trajectories on the browser. And here I define three spanning across a spectrum of agenticness. From least agentic, we have purposebuilt browser trajectories that are specially made to complete a single task. To the most agentic, we have just in

03:38 time browser automations where we don’t know what the task is beforehand and the user provides us an arbitrary task to complete. In the middle, we have something sort of between both. We’re using the browser as an implementation detail to serve a larger goal, such as performing deep research or competitor analysis. At scale, I really want to hone in on these transactional workflows. Not because it makes things easy, but that’s because we typically see at scale. These workflows are transactional by nature

04:11 where we want to complete a unit of work end to end. On the other side, browser agents provide us a flexible execution environment to handle ambiguity in production. So, we know how agents interact with the web and the different types of trajectories we may see and we’re honing in specifically on transactional workflows. Why is it so hard in production? And I think it comes down to a key interaction.

04:41 Cost accumulation is continuous while value realization is terminal. Each step in your trajectory, you’re incurring some incremental cost. Additionally, you’re incurring some incremental risk of failure. Yet, I only receive value from this browser trajectory when all steps of the task have been completed. That’s all to say that with browser agents, there is no partial credit. And the vast majority of web automations fall into this category. So you say, “Okay, well each step

05:13 there’s more risk and there’s more cost. If I use a highly capable model, I can sidestep this conundrum.” Take this thought experiment for an example. You have a browser agent where each step has an independent success rate of 99%. If your browser trajectory spans 100 steps, the overall success rate at scale starts to look like 36%. only about a third of the time. That’s not super great and it’s certainly not how we deliver reliable outcomes for

05:44 customers in production. And so you say, okay, well, we’re continuously accumulating costs, we’re continuously accumulating risk, and the percentage chance of value realization is very low. That would suggest to never use a browser agent. Uh, thankfully, that’s not the subject of my talk today. I think this dynamic is real, but the math isn’t the whole story. And so, it’s time to take a step back. Let’s think about what’s important and how agents actually

06:17 deliver value. And to illustrate the value that I wish agents to deliver, uh, I will propose a simple equation. Um, perhaps not used enough by many of us. And this equation is profit equals revenue minus costs. Yes, very simple. Um, but perhaps overlooked when thinking about browser agents. And this is all to say that your agent is just another line item and that your agent should give you more than it takes. But when we think about what your

06:48 agent should give you, that begs the question of what does a browser agent actually give you? And so to answer this question, I like to think about the context of the customer. At browserbase, we help our customers deploy browser automations for their users. Oftent times, they’re not automating the web for their own needs, but instead performing actions on the web on their customer’s behalf. And so, a browser agent actually gives you a task completed for your customer repeatedly.

07:19 Well, it gives you that ideally. And the question is how you can create a dynamic system that affords the benefit of being able to wade through ambiguous problems while also completing the task repeatedly every time when the environment stays the same. And so we’ll look across three dimensions and in order of importance. First is performance. Second is cost and third is maintainability. And when we think about performance, this is the primary measure

07:50 of success that we want to focus on. Because once we know a browser agent is performant, it can complete the task we ask it to do. Then cost and maintainability simply become optimization problems. And I like optimization problems because those can be addressed with engineering. So the first thing we need to ask is what success actually looks like. And success in these transactional browser trajectories look like some kind of concrete artifact that the run leaves behind. If you’re paying a bill, that

08:23 might look like a confirmation email. If you’re placing an order, that may look like an order ID or a receipt. Or if you’re submitting a form, that may look like a new record in a system that you can deterministically query. Second, >> how often is an agent correct? And there’s two ways to measure that. Naively, you may measure the success of an agent on a perr run basis. Assume you have a browser agent. It’s probability

08:54 of success on any run is 50%. Yet, if you permit that agent to retry for a particular transaction, maybe up to four retries, your per transaction success rate quickly climbs to 94%. And so when you think about what metric matters, again, I’d implore that we ground ourselves in the customer. The customer does not care necessarily how many retries you perform, but rather that the workflow was completed successfully and reliably. In short,

09:26 permit retries and measure success against a per transaction basis. >> Next, we’ll explore cost and we’ll break cost down into two dimensions. The first is per run, which today is predominantly model costs. But as engineering evolves in the open source landscape for models, there’s an incredible downward price pressure on the cost of intelligence. And I believe over time, model cost will become a smaller and smaller component

09:57 of deploying browser agent trajectories. Next is infrastructure and compute. What are the real resources you need to run your browser agent trajectories? And last is additional integration and tooling. Then we think about maintenance over time. We need to invest in observability not only in how the agent made decisions but what actually transpired in the browser. Second is developer time. When things break or things are not working optimally, people need to come in and reconfigure the

10:28 harness or augment its capabilities. And last is re-evaluation. As the problem changes, sites change, models progress, and we need to re-evaluate our point in time solution to last the test of time. So, we talked about performance, we talked about unit costs, we also talked about maintainability. What are the risk factors to the durability of these over time? Well, first and foremost, the

10:59 environment can fight back. Antibbots and just generally we understand the web as not the friendliest place for your agent. Second, the task changes. The underlying job shifts beneath the a shifts beneath the agent and we ask the agent to wade through ambiguity. Luckily, this also benefits from increased model capability as models are more able to intuitively understand the intention of the user. and therefore in the face of something go wrong course correct and ultimately complete the

11:30 goal. Third is a better methods better methods appear and this is sort of a canonical conundrum with automating the web. What happens if I’m automating the web today and that same website creates an API tomorrow but for the vast majority of browser agent use cases uh an API isn’t coming anytime soon. And finally the model strays off tasks. models are inherently indeterministic. So from run to run it may not ide follow the ideal path or the critical path to

12:01 perform the action. So finally we’ve come to the meat of the talk. How do we concretely generate shareholder value? This is the question. Uh we understand that if a model performs well then cost and maintainability are an optimization problem. So let’s explore that in practice together. We’re going to go through a highle architecture of a browser agent for real world tasks. This task is automating an health insurance portal. Every step we take in creating

12:34 this dynamic system should influence one of these key levers. Cost, performance or maintainability. So first we’ll start with a lowly agent. The goal of this agent is to log in and download an explanation of benefits. A request comes in, the agent performs the action on the browser, and this is the minimum necessary system you need to just do something. Next, you realize that when you go to download that explanation of benefits,

13:04 that downloading involves programmatically interacting with the page. And then once you perform that download, you need to retrieve that file from storage and bring it into the agent’s runtime. So you create a tool that does this all in one transaction. Instead of the model needing to make decisions on the fly, you encapsulate a complex operation into a single tool call. Next, you want to give the agent an opportunity to verify its work real time. This can be done by giving it

13:34 another tool. This tool is an OCR tool that deterministically extracts entities from the document and compares that against the system of record. Great. Now the agent knows if it’s did it if it really did its job right or not because it has a deterministic mechanism to verify that. Next, you realize that while this portal may have ambiguity in your business logic that the authentication workflow is never changing and you want to reuse the

14:05 same authentication mechanism for many business operations on the same portal. So you pull that authentication logic out from the responsibility of the model. Again, we’re reducing the number of steps that the agent has to take to complete the task successfully. This means lower unit cogs, higher performance, and easier maintainability as well as we can reuse this authentication function across multiple business logic workflows on the same portal. Last,

14:37 we’ve contained the responsibility of the agent to only the area of ambiguity necessary. So, we draft a skill. These skills look very similar to a standard operating procedure that we would have humans use in the past, except this standard operating procedure is for an agent, and it allows the agent to follow the critical path more closely. At the beginning of the talk, we discussed how the output of an agent can be thought of like a probability distribution. So if

15:08 we introduce the critical path to the agent, it knows it’s at step two and knows that it needs to take step three because we have outlined it in this skill, then it’s more likely to take that step three. It knows exactly where it needs to go and we remove ambiguity from the agents decision process. In the end, this is the final result. We have actually a pretty complex system. A request comes in, a serverless authentication function authenticates the browser. That browser is provided to

15:39 the agent runtime. The agent navigates the portal using the skill provided, calls a deterministic download function, provides that downloaded file to a verify tool, and can know authoritatively whether its trajectory succeeded or failed. In the end, this system takes steps out of the model’s responsibility that simply do not need to be in the responsibility of the agent. We reduce the number of steps it needs to take while still being able to achieve the

16:11 desired outcome. Thus, the system’s more maintainable and it’s more performant and we can overcome the browser agent conundrum. So, >> here we are. Welcome to production and congrats. Now you’re on call. Thank you.

16:46 >> [music]