Augmented underwriting: Apollo's AI-human partnership

When James Slaughter pitched new technology to skeptical underwriters at Apollo, he didn’t lead with efficiency metrics or cost savings. “What I want to do is I want to supercharge you. I want to take you from being great, to being the best in the world at what you do”, explained Slaughter on a recent episode of the Underwriting Intelligence podcast. This approach puts enhancement, not efficiency, at the heart of underwriting transformation and innovation.

That reframing is the driving philosophy behind augmented underwriting at Apollo, a specialty Lloyd's insurer that deployed 19 AI models against an original goal of seven or eight. Rather than viewing artificial intelligence as a replacement for human judgment, Apollo treats it as an amplifier.

What augmented underwriting looks like in practice

Augmented underwriting is a model where technology and human expertise work together rather than compete. Instead of technology telling underwriters what decision to make, the role of technology is to show underwriters past outcomes and ask questions that can help them reach an optimal decision: Have you made similar decisions before? Is this consistent with you previous choices?

Central to this is what Slaughter calls "the human in the feedback loop," a model where machines learn alongside humans rather than independently. The machine improves, but so does the underwriter, creating a cycle of decision intelligence that strengthens both over time.

This matters particularly in the London market, where human relationships drive business. The empathy, friendships, and professional bonds built over careers cannot be replicated by algorithms, and specialty insurance remains a fundamentally relationship-driven business even as digital tools reshape workflows.

Why trust is the real barrier to AI adoption

Fear drives much of the resistance to AI in underwriting. Experienced professionals initially believed technology initiatives aimed to eliminate their jobs, and that perception created an adoption barrier no technical sophistication could overcome. Only when the conversation shifted to augmentation rather than replacement did progress accelerate. Slaughter also notes the industry's tendency to jump ahead, pointing out that discussions about advanced concepts like agentic AI were happening before basic AI implementation was even complete.

As Slaughter explains: "If we're going to ask underwriters to use these tools, they've got to trust the output because there's no point if they sit there and go 'the output is not what I'm expecting. I don't trust it.' They're never going to use it."

Building that trust required credible data that's accessible, timely, and repeatable. Apollo invested 2-3 years focusing on data infrastructure before deploying advanced models, a commitment that proved essential. The payoff showed in deployment velocity: Apollo set realistic goals of eight or nine models but delivered 19 in the same timeframe once trust was established.

The three pillars of augmented underwriting

Apollo's approach rests on three pillars that address both business outcomes and the human psychology of adoption.

Performance: better decisions

The first pillar focuses on improving decision quality by putting insights in the right place at the right time. Apollo's risk scoring operates at both the per-risk level and the portfolio level, helping underwriters evaluate specific submissions while revealing patterns that might otherwise go unnoticed.

Apollo also implements what it calls non-pricing risk validation: a layer of contextual information beyond the technical price that accompanies their decision tools. This surfaces factors like historical consistency and portfolio alignment, allowing underwriters to see whether their current choices match past decisions while retaining full authority. The system doesn't prescribe decisions; it enriches the context surrounding them.

Productivity: smarter resource allocation

The second pillar addresses how underwriting talent gets deployed. Senior underwriters represent the highest compensation per unit of underwriting effort, so they should work on the hardest, most complex risks where their expertise creates the most value.

Without productivity tools, experienced underwriters might spend time on straightforward $50,000 risks simply because a familiar broker submitted them, while complicated submissions requiring senior judgment sit waiting. Augmented underwriting enables intelligent work distribution that also serves as a training and development tool, where junior staff gain exposure to increasingly difficult risks. In a hybrid working environment where traditional mentorship models have shifted, this systematic approach to professional development becomes especially important.

Engagement: taking value to market

The third pillar addresses how Apollo engages customers and brokers. Better insights enable faster response times and higher-quality products, and Apollo's smart follow capability, developed through broker partnerships, enables rapid, consistent responses that strengthen market relationships.

Brokers respond better to a "no and here's why" than a simple rejection. When underwriters can articulate the specific reasoning behind their decisions, backed by data and clear logic, the conversation shifts from frustration to understanding. When Apollo needed to exit certain risks during a challenging property market, clear communication backed by data made difficult conversations easier, and the logic behind those decisions, delivered proactively, turned rejection into relationship building.

Portfolio repositioning in practice

Slaughter shared a concrete example from early in his Apollo tenure. Facing a challenging property market with the Texas winter storm creating significant losses, the team developed a prototype of their risk scoring approach to quickly reposition the portfolio.

The approach was deliberately simple: four calculated fields from raw data, compared against decision metrics like price and capacity. Even basic scoring revealed that risks appearing equivalent actually differed meaningfully when examined through additional lenses, and ranking them enabled rapid prioritization during the busy Q2 property season. The team identified which submissions deserved immediate attention and communicated early with brokers about exit decisions, providing reasoning at every step.

The result was that Apollo repositioned its ENS property book and gained visibility into when to lean back into the market. The company moved aggressively a full year before competitors because the metrics revealed the opportunity. Now facing a property market where rates are declining, Apollo plans to repeat the exercise with far more sophisticated tools.

Building the data foundation

Apollo benefited from a greenfield environment on an Azure platform, meaning data and toolkits communicate natively. The organization can move in and out of workflow processes without compromising data quality or accessibility.

The 2-3 year foundational investment wasn't optional. As Slaughter emphasizes, trust only comes with credible data that's accessible, timely, and repeatable, so underwriters receive consistent information exactly when they need it. Without this foundation, even sophisticated models produce outputs that underwriters dismiss.

Cloud-native architectures provide substantial advantages for insurance technology deployment. Organizations built on modern stacks avoid the legacy system reconciliation that consumes resources at older firms. Legacy organizations can still succeed, but they require different approaches, realistic timelines, and often a different mindset about how quickly transformation can occur.

Leading adoption from the top

Successful adoption also requires specific leadership behaviors. Leaders must articulate a compelling vision, not by predicting the future with accuracy but by painting a picture that inspires action. Organizations need strong storytelling to bring colleagues and partners along, and leaders must empower teams to move fast and convert vision into tools underwriters can actually use. Quick wins matter because underwriters get bored if tools don't reach their hands fast enough, and perfection becomes the enemy of progress.

How hyperexponential supports augmented underwriting

Apollo's three-pillar approach aligns closely with how the hx platform is structured. For the performance pillar, Decision Engine gives actuaries a Python-native environment to build and deploy models that feed risk scoring directly into underwriting workflows. For productivity, Submission Triage uses intelligent ingestion to route the right risks to the right underwriters, so senior talent focuses on complex submissions rather than routine ones. And for engagement, Pricing & Rating puts decision context, from indicative quotes to portfolio impact, in front of underwriters at the point of decision, enabling the kind of fast, reasoned broker responses Slaughter describes.

The result is a platform that treats augmented underwriting as a connected workflow rather than a set of disconnected tools. When models, triage logic, and portfolio intelligence live in a single environment, every underwriting decision captures data that improves the next one. See how the hx platform connects triage, pricing, and portfolio intelligence in one workflow.

Hear James Slaughter share the full story of Apollo's augmented underwriting journey, including practical insights on building trust with underwriters and accelerating AI deployment. Watch the full podcast episode here.

Frequently asked questions

What distinguishes augmented underwriting from automated underwriting?

Automated underwriting removes human decision-making entirely. Augmented underwriting keeps humans central while providing better information and enabling faster workflows. The underwriter retains full authority over every decision.

How long does implementation typically take?

Timelines vary significantly based on organizational readiness. Companies with clean, accessible data can move faster, while those requiring legacy system integration need patience. The critical factor is whether leadership commits to foundational data work before pursuing advanced applications.

Does augmented underwriting work for all insurance lines?

The approach proves particularly valuable in specialty insurance where human judgment and relationships remain essential. More commoditized lines may benefit from higher degrees of automation, though the augmented approach still applies wherever underwriter expertise adds value.

What's the biggest barrier to adoption?

Underwriter trust. Technical infrastructure matters, but psychology matters more. If underwriters believe the output doesn't match their expectations, they simply won't use the tools, regardless of how sophisticated the underlying models are.

How do you measure success?

Apollo tracks deployment velocity, model adoption rates, and underwriter engagement alongside traditional performance metrics. Deploying 19 models versus a planned eight demonstrates how trust accelerates outcomes once the foundation is in place.

What skills do underwriters need to work effectively with AI tools?

Underwriters don't need technical skills to benefit from augmented underwriting. The value comes from domain expertise, relationship management, and judgment, which are the qualities AI tools are designed to amplify. What matters most is openness to using data-driven insights alongside experience, not replacing one with the other.

How does augmented underwriting differ from submission triage?

Submission triage is one component within an augmented underwriting approach. Triage focuses specifically on prioritizing and routing inbound submissions, while augmented underwriting spans the full decision workflow: from risk scoring and pricing validation to portfolio-level analysis and broker engagement.