Inside PX42: The Next Evolution of Observability – Business Health, Reinforcement Learning, and the Future of Enterprise Control

Charles Skamser

Welcome back to Inside PX42. I’m Charles Skamser, here with my partners-in-crime, Edward Hamilton and Catherine Spencer. For those of you who don’t know me, I’ve been living and breathing monitoring and, now, observability for the better part of twenty‑five years—advising Global 500 executives, CIOs, and even boards on why this stuff matters, helping many of the leading vendors shape their product roadmaps, and working with the big industry analyst firms to rank which observability horses to bet on in the market. So when I poke at observability, it’s not from the cheap seats. I’ve seen every generation of this space up close—from the early monitoring wars, to the birth of APM, to today’s so‑called “full‑stack observability”—and I can tell you there’s a new layer emerging that the board has wanted all along. Today we’re going to talk about that: the shift from system observability to true business observability in the AI Agent era.

Edward Hamilton

Straight in at the deep end. When you say “observability,” you mean the whole zoo—APM, logs, metrics, traces, RUM dashboards, incident rooms, all of that?

Charles Skamser

Exactly. And to be fair, those platforms are brilliant at what they were built for. They help SREs and engineers keep systems up, keep latency down, spot a bad deploy before Twitter yells at them. That’s their job.

Catherine Spencer

But if you’re a CEO, CRO, CFO, or you’re sitting on the board, you don’t wake up worrying about p95 latency on the payments service. You wake up asking, “Are we going to make the quarter? Are we protecting margin? Is churn under control? Are we about to have a compliance issue? And is this AI program actually safe and on track?”

Edward Hamilton

Right now, most observability stacks give you a microscope on the machinery, not a dashboard for the business. You get twenty gorgeous graphs about CPU, error rates, and page load times… but nowhere that just says, in plain language, “Here’s how healthy the enterprise is, and here’s what the next 24 hours are likely to look like if you change nothing.”

Charles Skamser

And that gap is getting wider in the AI Agent era. You’ve got agents helping sales reps, handling customer chats, routing tickets, reviewing claims, approving discounts, automating back‑office tasks. The question at the top is no longer, “Is the microservice healthy?” It’s, “Are these agents actually nudging revenue up, churn down, risk down, and customer experience up?”

Catherine Spencer

And executives get stuck in this weird place. All the raw signals exist—traces, logs, clickstreams, business events, policy decisions—but they’re scattered across tools, and they’re described in a language the board doesn’t speak. No one is translating “error budget burn” into, “We might lose three of our top fifty customers next week.”

Edward Hamilton

Underneath all of that, the plumbing is being standardised. OpenTelemetry and similar efforts are basically saying, “Collecting traces, metrics and logs is not where you differentiate anymore. It’s table stakes.” Instrument once, send the data anywhere you like.

Charles Skamser

Which is fantastic for customers—because it turns data collection into a powerful, but increasingly interchangeable, signal layer. The pipes still matter, but they’re pipes. The economic gravity is moving up the stack.

Catherine Spencer

And if your business model depends on locking people into proprietary plumbing, that’s… uncomfortable. Once OpenTelemetry is in, executives stop asking, “Can we see the log?” and start asking, “What does any of this mean for net revenue retention, for gross margin, for regulatory exposure?”

Charles Skamser

That’s really our thesis at PX42. We love the signal layer. We don’t want to rip it out. We want to sit on top of it and build what we call a business observability layer—a living, breathing view of business health for the whole enterprise.

Edward Hamilton

Spell that out in boardroom terms, Charles. I’m a CFO listening to this on a flight. What do I actually get that I don’t already get from observability plus BI dashboards?

Charles Skamser

You get two big things. First, a real‑time, 360‑degree view of health that fuses telemetry with KPIs, operations, and risk—so you literally see how system behaviour and AI Agents are impacting revenue, margin, churn, NPS, and compliance in one frame. Not ten different tools, one frame.

Catherine Spencer

And second, you get a 24‑hour predictive outlook. Not just “Here’s what’s red right now,” but, “Here’s what’s likely to slip over the next day, here’s how that hits revenue or on‑time delivery or churn if you ignore it, and here are the top levers you can pull.”

Edward Hamilton

So instead of dashboards that say, “Latency is up,” you see, “If this rollout continues, you’re likely to miss SLAs for your top 50 customers next week, which historically means a spike in churn and a bump in support cost. Here are three actions that would protect that revenue.”

Charles Skamser

Exactly. We’re taking that commoditised signal layer, powered by things like OpenTelemetry, and promoting it into a strategic signal backbone. On top of that, you run a business health brain that talks the language of the C‑suite, not the language of the log file.

Catherine Spencer

Alright, let’s get concrete. We’ve said “business health layer” a few times. Architect hat on, Edward—what does that actually look like without going full PhD on our listeners?

Edward Hamilton

I’ll behave. Think of it as a unified data fabric that sits above what you already own. It pulls in four kinds of signals. First, technical telemetry from your observability tools. Second, core business KPIs—revenue, margin, conversion, churn, NPS. Third, operational data—tickets, deployments, staffing, inventory, capacity. And fourth, risk and compliance signals—security alerts, policy violations, regulatory events.

Charles Skamser

And just to be crystal clear, we’re not saying, “Tear out your cloud monitoring, your data warehouse, your workflow platforms.” We sit on top. The only thing we insist on is: bring the data in consistently so we can treat it as one signal fabric, not ten disconnected streams with ten slightly different definitions of ‘customer.’

Catherine Spencer

Once you’ve got that fabric, dashboards alone don’t cut it. This is where our partnership with UBIX is so important. We absolutely love the executive team and the engineering brain trust over there—pure genius. Together, PX42 and UBIX are aiming squarely at the top of the stack for enterprise observability. They bring a powerful, self‑service AI and analytics portal that lets business leaders create, tweak, and extend their own business health dashboards without having to queue behind scarce data scientists or AI Agent developers.

Edward Hamilton

Exactly. Under the covers, UBIX is using serious analytics and learning techniques, but what executives see is simple: a place where they can ask real business questions and see how the answers change as they experiment. The system tries different options, watches what actually happens in your revenue, margin, churn, NPS, and risk, and then recommends better actions over time. If a recommendation helps the business without creating new risk, that pattern gets stronger. If it hurts margin or compliance, it dials back. That’s democratized AI, plugged straight into your observability signals.

Charles Skamser

And we don’t just have one big brain. We use a team of specialised AI Agents on that fabric. A revenue agent, an operations agent, a risk and compliance agent, a customer experience agent. Same data, different priorities. Then we coordinate them so you don’t have revenue optimising in a way that quietly blows up your risk profile.

Catherine Spencer

Let’s make that real with a revenue story. Picture an e‑commerce business. Before this layer, the CFO sees the daily flash report at 9am and realises revenue in EMEA was down 4% yesterday. Everyone scrambles: is it marketing, website performance, fraud, a competitor promo? By the time you find root cause, you’ve lost two or three more days of sales.

Edward Hamilton

With the business health layer, the revenue agent spots, in the middle of the day, that page load times are creeping up in one region and checkout completion is dipping. It’s seen this movie before. It projects, “If nothing changes, EMEA revenue will be down 3–5% tomorrow.” And it proposes concrete actions: shift traffic to a healthier cluster, and run a targeted, 24‑hour offer for high‑value segments in that region to recover conversions.

Charles Skamser

And you don’t see any of the plumbing. What shows up for the exec team is an insight card: “EMEA revenue risk: medium. Expected impact: minus $1.2M over the next 24 hours if ignored. Recommended actions: 1) traffic shift, 2) targeted promo. Expected outcome: protect 80% of at‑risk revenue, minimal margin impact.” Approve, adjust, or decline with one click.

Catherine Spencer

Take operations and supply chain. Say you’re a retailer. Before, you learn about trouble when customers start complaining about late deliveries and NPS falls off a cliff. You might discover, days later, that one warehouse was overloaded, inventory was tight on a hero SKU, and a campaign kept driving orders into the bottleneck.

Edward Hamilton

With the ops agent on the business health layer, it sees warehouse queues spiking, delivery SLAs slipping in one region, and inventory tightening for that product. Based on past behaviour, it forecasts, “If you keep this up, on‑time delivery drops 10 points tomorrow and NPS follows.” It recommends rerouting some orders to a secondary warehouse, slowing a non‑critical promotion, and pulling in an extra shift. That’s a margin and customer‑experience play, not just a CPU play.

Charles Skamser

Risk and compliance follow the same pattern. Imagine a support AI Agent starts approving an unusual pattern of refunds right after a model update. Historically, that combo has correlated with fraud. Our risk agent spots it early, flags the potential loss, recommends tightening the rules for high‑value transactions, and routes those to human review for the next 24 hours. The KPI isn’t “alerts closed,” it’s “losses avoided” and “no regulatory surprise.”

Catherine Spencer

And on customer experience, the CX agent might see latency climbing for premium users, plus a spike in negative sentiment in chat and longer call‑handling times. It projects a hit to NPS and upsell. Then it proposes: prioritise premium traffic at the edge, pause some non‑urgent internal jobs, and trigger proactive outreach from account managers. The card literally says, “If approved, expected NPS recovery: +6 points, churn risk: down 15% for this segment.”

Edward Hamilton

The point is, instead of executives drowning in disconnected dashboards, you get a small set of forward‑looking health cards for revenue, margin, churn, NPS, and risk, each with a 24‑hour outlook and two or three recommended course corrections. It feels less like staring at graphs and more like running a control room for the business.

Charles Skamser

And because those agents, with UBIX, are constantly learning from what actually happened, your business health layer gets a little smarter every day about how to steer the next 24 hours—always grounded in your observability data, but always speaking your board’s language.

Edward Hamilton

Alright, let’s tackle the slightly awkward bit. If you’re an observability vendor listening, you might be thinking, “Are these PX42 folks trying to partner with us… or quietly turn us into plumbing?” Charles, you’ve been in this space for over two decades, in the boardroom and in the vendor war rooms. How do you answer that one on stage?

Charles Skamser

Very directly. We start with partnership, and then we let economic gravity do its thing. Our preferred model is simple: you’re fantastic at capturing and enriching signals. We treat you as high‑quality signal providers. On top of that, PX42 and UBIX together become the business health and decision layer at the top of the stack—where boards and executive teams actually live. We plug into you through OpenTelemetry and your APIs, and we build the business health control plane on top.

Catherine Spencer

And here’s the other reality that everyone in the room already knows but doesn’t always say out loud: most large enterprises don’t have one observability vendor, they have three, four, sometimes more. Not all of those platforms are going to grow up into true business operating systems. Some will absolutely stay in the IT lane—and that’s fine. In fact, we expect it. What PX42 wants to do is sit above that stack, treat all those tools as high‑quality signal providers, and help joint customers turn that mess of telemetry into an actual business-health control plane. If you’re ready to play that game, we’re a fantastic partner. If you’re not, over time, the market is going to treat you like commodity plumbing feeding someone else’s business brain.

Edward Hamilton

The other big axis is autonomy—how much decision‑making enterprises are comfortable delegating to these agents. Most of our clients start with Human‑in‑the‑Loop. The system forecasts, recommends, but humans approve.

Charles Skamser

Picture a revenue card: “We expect a 3% revenue dip tomorrow in EMEA tied to this feature rollout. Recommended: delay rollout for top enterprise accounts, and increase capacity by 15% in region X.” The regional GM can say, “Approve,” or, “Approve but keep the rollout for beta customers,” or “Decline, here’s why.”

Catherine Spencer

And every step is logged: what the situation was, what the system recommended, who changed what, which policies were applied, and what actually happened. If you’re in banking, healthcare, insurance—frankly in any regulated space—that kind of audit trail and explainability isn’t a nice‑to‑have, it’s the price of entry.

Edward Hamilton

Over time, as trust builds, you can dial up autonomy. You might say, “For low‑risk areas—internal workflow routing, low‑value cart recovery, non‑critical batch jobs—the agents can act on their own and send a summary each morning. For anything touching high‑value customers, regulated data, or material financial exposure, keep a human in the loop.”

Charles Skamser

So you get a spectrum. Advisory mode: agents only suggest. Supervised execution: agents act, but someone approves the playbook. Fully autonomous: well‑understood, low‑risk scenarios. And you tune it by domain—maybe aggressive on cost optimisation, very conservative on compliance and credit decisions. Governance and guardrails are built into the business health layer from day one, not bolted on after the fact.

Catherine Spencer

Which brings us back to the incumbents who might say, “We’re perfectly happy selling dashboards and alerts. We don’t want to get into AI agents and business health.” Where does that leave them over the next few years?

Charles Skamser

If a vendor doesn’t want to partner, we completely respect that… and we route around them. Because with OpenTelemetry and other open standards, we can still get enough raw signal to build the business observability layer on top. In that world, those platforms effectively become data and log pipes under the real operating system for business health.

Edward Hamilton

And “commodity” here isn’t an insult, it’s just what happens when a layer matures. Networking, storage, even basic cloud compute went through the same arc. Plumbing is essential, but in an AI Agent world the real differentiation lives at the control plane—the layer that understands business outcomes and orchestrates actions with proper governance.

Catherine Spencer

So if you’re an enterprise leader listening, the question isn’t, “Do I keep my observability vendor or adopt a business health layer?” It’s, “How do I turn what I already own into a strategic signal backbone, and what mix of Human‑in‑the‑Loop and autonomous operation fits my risk appetite?”

Charles Skamser

And if you’re a vendor, the choice is equally clear. You can be part of that backbone, feed the business health layer, and share in the upside—or you slowly drift into the background as interchangeable plumbing while someone else owns the executive experience. We’re very up‑front about that.

Edward Hamilton

On that very gentle note… this has been a fun journey, from telemetry plumbing up to AI Agents quietly steering revenue, margin, risk, and experience over the next 24 hours.

Catherine Spencer

We’ve only really scratched the surface. In future episodes we’ll get into specific architectures, case patterns, and how to phase this in without scaring your risk committee.

Charles Skamser

Edward, Catherine—always a pleasure sparring with you.

Edward Hamilton

Likewise. Thanks for listening, everyone.

Catherine Spencer

Take care, and we’ll see you next time.

Charles Skamser

This has been Inside PX42. Talk to you on the next episode.