What's Up with Tech?

Beyond Copilots: Agents That Do The Work

• Evan Kirstel

Share episode

Interested in being a guest? Email us at admin@evankirstel.com

Copilots can suggest the next click, but they rarely deliver the finished job. We dive into a different path: enterprise AI agents that integrate with your systems, understand your business rules, and execute end-to-end workflows with governance, accuracy, and reliability. Rob Bearden co-founder and CEO Sema4.ai shares how their platform moves beyond brittle scripts and UI macros to reasoning-driven automation that adapts to changing contracts, policies, and supply constraints—turning strategy into repeatable, measurable outcomes.

We trace the journey from big data to autonomy: insights and KPIs used to point the way, but humans still had to do the work across dozens of apps and tabs. Agents close that last-mile gap by reading documents, joining data across ERPs and CRMs, and following rule-bound reasoning paths to finish the task. You’ll hear concrete wins like multi-page invoice reconciliation done in minutes with higher accuracy, AP help desk cases resolved without swivel-chair searches, and quote-to-cash automated across fragmented systems. The result is less toil, fewer errors, and outcomes you can audit and scale.

If you’re stuck in AI pilot purgatory, the way out is a platform strategy and tight guardrails. We break down a crawl-walk-run approach: pick a high-leverage use case, define the outcome, run a focused proof, measure ROI, then rinse and repeat. We also scan the broader agent ecosystem—Salesforce, ServiceNow, and hyperscalers are leaning in—while making the case for an enterprise-wide layer that spans SaaS apps, data warehouses, and data lakes. Finance operations lead early adoption, but healthcare, insurance, and manufacturing are close behind, wherever people juggle multiple systems to make a decision.

Ready to trade tab hell for trained agents and predictable outcomes? Follow the show, share this episode with your ops and finance leaders, and leave a review to help more teams find it.

Support the show

More at https://linktr.ee/EvanKirstel

SPEAKER_00:

Hey everybody, fascinating chat today with a company who says co-pilots are yesterday's training wheels. Uh looking forward to chatting with Rob from Semaphore.ai. Rob, how are you? I'm doing great. Thank you for having us today. Well, thanks for being here. Really excited for this chat. How would you describe your vision and mission at Semaphore these days?

SPEAKER_01:

Well, we're very excited about the overall market and the embracement around AI agents and the role that we play with our enterprise customers and how we help them build, run, and manage end-to-end very uh deterministically driven uh complex workflows.

SPEAKER_00:

Fantastic. And you know, your AI agents, AI agents in general, super exciting, but how do they differ from today's co-pilots or let's say even professional AI systems?

SPEAKER_01:

Sure. So, you know, our agents are we we enable our environment to integrate with our customers' internal systems and they understand the business context that our customers are trying to get an outcome for. And then we autonomously execute very complex end-to-end workflows. And really, unlike co-pilots, which really just assist and rely heavily on UI interactions, our agents perform the work in an end-to-end manner with high accuracy governance and enterprise grade and reliability, all built into the outcomes that we enable for our customers to accomplish.

SPEAKER_00:

Fantastic. Love the outcome focus. And you know a thing or two about outcomes. You've helped scale companies like JBoss, Fortin Works, Docker, and others. So, what made you take the leap into a venture like semaphore.ai?

SPEAKER_01:

Well, and in our previous uh companies, we've been, we we came out of the big data world. And the big data world was about bringing data under management and then enabling use cases that got to machine learning, and then that advanced into predictive analytics. And the big air pocket was being able to go from having insights about a best practice or um leading KPI to actually being able to execute the outcome through the work. And that's where the big opportunity is with AI generally and agents specifically is how do you execute work in an autonomous way that leverages data in place and drives an outcome autonomously in a best practice manner for a customer. And that's why we started Sumaph4 AI was to provide a platform our customers could build highly reliable, safe agents to execute world-class outcomes.

SPEAKER_00:

Fantastic. I was at Gartner's symposium a few weeks back, and there was one insurance company on stage talking about 74 different AI pilots they were working on. And so many enterprises get stuck in this sort of AI pilot purgatory. Uh, why is that? And what how do you what do you see as the uh way out?

SPEAKER_01:

Well, I think I think to your point, there's a lot of experimentation in general happening around AI. And that experimentation comes in different form factors, everything from what's the art of the possible to can it accomplish this use case idea that we have to where and how do we apply AI into our IT stack? And so the common denominator behind either constrained or even failures is there's lack of an objective, there's lack of a strategy. And what we bring is a very clear path for our customers to be able to execute very clear, concise, compelling outcomes through agentic work. And so we give them the guardrails to enable outcomes in a very predictable, consistent, safe way, um without having to go through a guessing exercise uh and and and uh hope that they'll get an outcome. We give them those guardrails to get to very deliberate, precise outcomes.

SPEAKER_00:

Very cool. And you know, companies have been using technologies like RPA for a decade, uh many are already deep in different automation projects, but what's the best entry point into an agent-based uh system like yours?

SPEAKER_01:

Well, I I think to your point, RPA was was really um a process that showed the art of a possible for getting to an outcome. And so I think that was um uh ground zero work, if you will. And what agentic does is instead of having just in the RPA world a point-to-point hard-coded outcome, true agentic gives you the ability to reason through uh a series of steps to get to a best practice outcome that reasons with documents and data that are required to deliver on the work. And so it's really a much higher, better outcome purpose that's outcome driven versus definitive point-to-point, because the challenge with a point-to-point outcome is that business rules change, business relationships changes, product and supply chain uh velocity and constraints change. And in RPA point-to-point world, they're constrained and hard-coded. In an AI agent outcome with our environment, we know how to reason to the outcome based on all the changes and constraints that evolve and accommodate for that with our platform.

SPEAKER_00:

Fantastic. So, my last real job was at Oracle over a decade ago. And at that time, you know, I had dozens and dozens of apps, homemade internal employee tools and apps, and uh you can imagine what that was like. And for many, it hasn't changed that much. You still got people with 100 tabs open and all kinds of things to navigate. How do you think AI agents will really change day-to-day enterprise workflows for your average employee or worker?

SPEAKER_01:

Well, I think the way we have enabled um our customers to embrace AI and enable AI in their environment is just as you train an employee to do that work you referenced, you train an agent to do that work. You define the outcome, and the agent executes that workflow. And that agent learns the reasoning paths that get to the most efficient, effective outcome as defined by the human. And then instead of training a human to do that, the agent continues to execute that, and uh and and and we get to very precise, accurate outcomes that are continuously learning and accommodating for the changing constraints that inevitably embrace a supply chain, a product cycle, a customer relationship, et cetera.

SPEAKER_00:

Fantastic. And do you have any anecdotes or stories around capabilities that agents are unlocking today, workflows or ideas that your clients have come up with that, you know, copilots can't deliver?

SPEAKER_01:

Yeah, well, uh again, I think our agents are doing things very differently than what a co-pilot would do, which is more of a personal assistant. Our agents are focused on executing work outcomes and delivering on the work in an autonomous way through the reasoning that's required to get to that best practice outcome. So, a couple of examples around that, where you're training in the world of invoicing, where you have a lot of multi-page, highly complex invoices that have to be reconciled to get to the proper payment or the proper adjustment based on either contractual uh achievements or milestones or discount achievements or milestones. And instead of humans having to go through and manually evaluate, reconcile, and adjust, we understand what those business constraints are, what the rules behind those reconciliations are, and the agents can do what takes usually many hours for humans to do in the matter of just minutes. And and more importantly, or as importantly, do it every time with highest degrees of accuracy, much higher than humans, in a more predictable automated way. Um another example, um AP help desk, um, to where our customers are very concerned about uh their accounts payable being done accurately and on time, when their customers have questions about their bill or their payment terms or the amount, as they log in for uh those questions to the AP department, humans have to go through in a manual way and research, evaluate across multiple systems, multiple documents, and then reason through what is the best remedy. And many of our customers have used our agents to actually automate that entire evaluation and outcome and reconciliation and payment process. Another great example is where some of our customers have had to go to, they have many quotes that they put out to customers, and they have to understand where they are in their quote to order to cash cycle, and they and they are having to go manually parse through that with inside sales teams, accounts receivable and payments teams. And many of our customers have actually automated that quote to order and order to cash cycle, leveraging our platform and our agents to do that.

SPEAKER_00:

Amazing. So talk about the AI agent ecosystem, if you would. A lot of uh early excitement and hype, of course. Um, companies like Salesforce potentially even changing their brand to Agent Force. You know, are we getting ahead of our skis, or do you see that 2026 will be a real list for the entire ecosystem, uh, solving some of the challenges like interoperability and interworking and all those things that are still early days?

SPEAKER_01:

Well, I I think to your point, whether it's Salesforce or ServiceNow or the hyperscalers, everybody has embraced solidly the future is going to be enabled by agents. And work should be able to be leveraged and delivered through an agent. Then the question becomes what's the highest and best value to an enterprise in an agent strategy? Certainly there's going to be point agentic capability through the point SaaS applications. You brought up Salesforce, but really every ISV that out there, every SaaS application provider has to be able to provide some form of agentic capability. What we're really seeing, though, in the bigger, broader movement is the enterprise is saying, look, we need to have an end-to-end autonomous outcome and a predictable way to do that, not just in a siloed application, but across a workflow process. And so we want to be able to have a platform strategy that we can define outcomes across the workflows that are interdependent across multiple systems, multiple data sets, whether they're transactional, whether they're DW, data warehouses, data lakes, uh, our SaaS applications, and to be able just to define the work and have an agentic platform that understands where the data and documents are that enable the agents to reason and execute the workflow autonomously. And that's what we're enabling the enterprise and the ISD community to do through our platform.

SPEAKER_00:

Fantastic. Are there any industries that are ripe for change and agent adoption? I mean, I think of healthcare where you know I go into my local doctor's office and I there's, you know, a couple people at the front desk and then about 10 people in the back shuffling paperwork. It's it's unbelievable. I imagine that industry and many others are in a similar boat.

SPEAKER_01:

Yeah, I know that I think that's the that's the the real paradigm shift that we're seeing across the industry is that it's not just concentrated in one industry and or line of business. It's actually across the workflows, whether it be financial services, healthcare, insurance, industrial manufacturing, et cetera, et cetera, et cetera. But it's it's the work where we're training humans to do very manual things across multiple systems that require access to data and documents to come to a reason decision. We now can do that autonomously with agents through our platform. And you know, great examples of that are certainly all the industries we just touched on, but if we look at where early use cases are being very successful, it tends to concentrate in the back office of the CFO, where there's high leverage that comes in areas like tax, compliance, HR onboarding, IT InfoSec, RevOps, those kinds of areas where there's a lot of humans that have to look at multiple from data across multiple systems of record that and and multiple document types in order to define the outcome that they're trying to get to and enable that outcome. We can automate that now through our agentic platform.

SPEAKER_00:

Fantastic. So it seems like there's going to be a big gap between enterprises of all sizes, those who adopt and jump on this agentic platform, and those who don't, who just continue to do things the way they've always done them, uh, which is most companies, I think. Uh what's the next step an enterprise or SME even can should take to avoid being left behind? And if if they were to work with you guys, what is the typical engagement look like? How long before you can see a return on the value and investment of time and more?

SPEAKER_01:

Well, uh, you know, the old saying innovate or die um sort of comes to mind in this new world of AI generally and agents specifically, and what the what the opportunity to get leverage and value capture comes from. What we focus on with our customers is uh crawl walk run. Pick a use case that creates value, do a proof of concept, prove that value very quickly, um define the outcome and the value capture that use case should create and generate, enable that use case, measure the outcome, rinse and repeat, and continue to drive through use case by use case, very, very uh maniacally focused on surgically extracting value from each use case based on the outcome definition that use case should deliver, and try to eliminate the experimentation and toil and be surgically focused on value creation.

SPEAKER_00:

Fantastic. Well, thanks for the insight into the vision, and it's an exciting one. Uh, where can people meet you in the new year in the spring? I guess you'll be out and about many places, customers, event, industry events, etc.

SPEAKER_01:

Uh that that we will be and be you know, we're maniacally focused on value creation and capture across our customers and uh um you know are are seeing great inflection points across the company. So we're excited to see what 26 brings and the acceleration of AI adoption and and and agentic enablement.

SPEAKER_00:

Fantastic. Well, thanks so much for sharing the uh the vision. Really exciting stuff. And thanks everyone for listening and watching. Also, check out our TV show, techimpact.tv, known Bloomberg and Fox Business. Thanks, everyone. Thanks, Rob.

SPEAKER_01:

Have a great day. Thanks for your time. Talk soon. Thanks. Bye bye.