Notes from
the field.

Growth doesn't break your strategy. Growth breaks your plumbing.

When you're small, everything runs on hustle and relationships. When volume doubles or triples, the seams rip in places nobody's watching. Three places I've seen it happen live.

Submission intake — always first. At Atlas General we tripled submission volume from under 10,000 to just under 30,000 annually in two years, all California workers' comp. At Kinetic, we were processing thousands of unique middle-market submissions a year averaging $200,000 each. If your underwriters are pulling submissions out of an email inbox, sorting PDFs by hand, manually setting up files — they're drowning before they ever get to make a risk decision. At Atlas we solved this with an offshore BPO team in India that set up every account in Duck Creek overnight. Our underwriters walked in each morning to files ready to work, not ready to organize. That one design decision is what let 10–12 underwriters handle nearly 30,000 submissions a year.

Quote timing — sneaky and structural. At a prior company I learned the hard way: release quotes 30 days before renewal and you've handed every broker a target to shop around. At Atlas we held quotes tactically until 10–15 business days before the effective date. That single shift moved our hit ratio from around 12–15% up to 15–18%. Timing is a competitive weapon most programs never think about.

Hunting versus farming — the trap. At Kinetic I recruited 30 of the Top 100 commercial brokerages in 90 days, before the first policy ever bound — relationships I'd built over years. By year-end we had over 65 appointed agencies and 400 branch offices producing. Getting them in the door was the easy part. Keeping them active and educated while simultaneously processing their submissions is the real challenge.

The punchline is simple. The first bottlenecks are never strategic. They're operational. Intake design, quote timing, and the capacity to actually serve what you've sold. If you can't process what you attract, growth works against you.

Let me put real math on it first, because the math is what separates hope from a plan.

To get to $100 million — depending on your average premium — you might need 33,000 to 40,000 accounts at small-commercial level, or far fewer at middle-market size. Either way: the operating model that got you your first $20 million cannot carry you to $100 million. Full stop.

So what breaks? Three things.

First, the founder's ability to touch everything. At Atlas I was the first distribution leader hired. The company had about 200 agencies doing roughly $20 million. Within two years we were at $150 million with 400+ agencies. The only way that happened was by building an inside marketing team, implementing agency scorecards, and performance-managing agencies with real accountability. 50-plus underperformers ultimately had to go. That sounds harsh. The reality is you cannot scale if the founder is still touching every deal. The math won't let you. I'm running the same play right now at Berkley Small Business — wholesale trading-partner network refocused from 42 to 24, fewer, deeper, better fit. The capacity you recover from underperformers is as valuable as a new hire.

Second, distribution outpaces underwriting capacity — unless you've designed your intake to absorb it. At Kinetic we hit $25 million in year one against a $7 million target — 357% of plan, and $45M in 18 months from a standing start. The underwriting team would have drowned if we hadn't built intake to handle that volume. The offshore BPO at Atlas was the design decision that made everything possible. The BPO handled setup. The underwriters handled judgment. That's a design choice, not a staffing decision.

Third, the data infrastructure isn't ready. At Kinetic we ran on Google Docs and email for 14 months — and our technical staff from the wearable-tech side maximized Google's capabilities further than most people thought possible. But by the time we crossed $45 million, we needed real pipeline visibility. We implemented Novidea and it was like putting on glasses for the first time.

The punchline: what breaks between traction and scale is the gap between the entrepreneur's instincts and the organization's systems. You have to institutionalize what used to live in one person's head. That's the bridge from $20 million to $100 million.

Loss ratio discipline and book mix. Activity feels good — submissions are up, the team is busy, the founder is on the road, the brokers are happy. But if average premium is dropping and the loss ratio is creeping, the program is consuming itself.

Scalable programs have written down what they will and will not bind. They have terminated underperforming agencies. They have a clear sense of which lines and which class codes their underwriting actually wins on. The non-scalable program is the one where the answer to "what's your target account profile?" is "we look at everything." That answer means no profile, which means no compounding.

At Atlas, the top 50 of our 400+ agencies drove the majority of written premium. The bottom 50 were on structured success plans, and 50+ ultimately had to go. The capacity you recover from underperformers is as valuable as a new hire. That math doesn't change at $20M, $100M, or $300M. I've now run that play across five builds — Pride Risk, Atlas, PacificComp/CopperPoint, Kinetic, and Berkley Small Business — cumulatively over $800M in GWP scaled. Scale belongs to the operators who run it, not the ones who chase it.

A few quick tells. The underwriters cannot tell you their hit ratio by submission source. The producer cannot tell you which agencies in their territory are above and below the line. The CEO cannot show you loss ratio by vertical for the last 12 months without an analyst pulling it. The CRM is half-populated because nobody enforces the discipline. Renewals are tracked in someone's email folder.

None of those are software problems. They are management problems. If the operating data isn't trustworthy, every strategic decision is being made on hope.

The companies that scale build the operating discipline first, then the technology supports it. The ones that get it backwards — buy the CRM expecting it to fix the management problem — end up with a more expensive version of the same chaos six months later.

I can answer this from three very different seats.

Seat one: the startup. At Kinetic I was employee #2 at a venture-backed MGU. Lean, fast, resourceful. We assembled a tech stack from best-in-class SaaS tools rather than building proprietary systems — Google Docs for pipeline tracking, InsurePro for policy admin, Zywave for customer data. That's smart when capital is precious — you preserve cash for underwriting capacity and distribution. We hit $45 million on that borrowed infrastructure before transitioning to Novidea.

Seat two: the well-funded MGA. At Atlas we had capital to do it right from day one. Duck Creek for submissions and policy issuance. Salesforce for CRM. And the offshore BPO team in India. Here's what made the BPO transformational — while we slept in California, they were clearing and organizing every file. Our underwriters walked in at 8 a.m. to ready-to-work queues. With 10–12 underwriters we processed nearly 30,000 submissions a year and grew from $20M to $150M. There is no world where those underwriters hit that volume if they were doing setup and quoting. The BPO handled setup. The underwriters handled judgment. That separation of duties is a capital-enabled design choice.

Seat three: the carrier turnaround. I joined PacificComp in December 2014 with the book at a combined ratio north of 200% — an Alleghany Corporation portfolio company in real trouble. The work wasn't a green-field build; it was a rescue. We took submission flow from about 8,000 in 2014 to 20,000 by the end of 2015, sustained around 25,000 a year through the rest of the run, verticalized into middle-market accounts $250K+ — restaurants, hotels, auto dealers, grocers — and got back to operating profit in 2016 and underwriting profit in 2017. The book scaled from $65M to $311M. Carrier capital let us run two engines in parallel: a straight-through-processing portal for small workers' comp — 100 eligible class codes (focused on the top 10 cleanly), up to $50K in manual premium, instant quote and bind — that grew from $250K to $32M over six years, managed by a single underwriter at a 25% in-force loss ratio. And a middle-market operation with 14 underwriters carrying the rest. Both worked because capital let us design the right architecture for each lane.

The bottom line. Capital doesn't determine success — it determines your design options. Startups must be ruthlessly efficient with borrowed tools. Funded MGAs can separate intake from underwriting. Carriers can build parallel tracks. The common thread across all three: use whatever capital you have to buy back your underwriters' time from administrative work.

When the manual workaround starts breaking trust.

If producers are losing accounts because submissions are sitting too long, the manual process has aged out. If underwriters are missing renewals because they're tracked in someone's email, you waited too long. The signal is not a dollar threshold — it's the moment retention starts slipping because operations stopped scaling.

At Kinetic we ran on Google Docs and email for 14 months — from a standing start to $45 million in premium — before moving to Novidea, because visibility had broken. Three months earlier would have been smarter. When we finally implemented Novidea, it was like putting on glasses for the first time. Most teams wait too long because the cost is visible and the breakage is not.

More than most teams measure, which is part of the problem.

The friction shows up as cycle time, as lost submissions, as average premium dropping because the underwriter doesn't have time to dig in. But it also shows up in places people don't measure: which submissions the underwriter chose to ignore because the PDF was a mess. The selection bias is real. The team thinks they're underwriting the book; in fact, the email format has been doing a quiet pre-selection for years.

At PacificComp we eventually shut down the submission email box entirely — no exceptions — and front-loaded vertical-specific intake questions. Auto dealers had their own set. Hotels had theirs. Grocers had theirs. That single decision shrank our clearance team from five people to two and eliminated the endless back-and-forth emails killing turnaround time. Three full-time equivalents of capacity recovered through design, not hiring.

When you finally put structured intake in place, the first surprise is what kinds of accounts you start binding that you weren't seeing before.

In my experience, about 70-30 in favor of the non-underwriting work. McKinsey puts the industry estimate at 30–40% of underwriter time on administrative tasks like rekeying data — and that's a charitable read.

Real risk evaluation — the actual judgment call — is maybe 30% of an underwriter's day in the best-run shops. The other 70% is finding the loss runs, asking the broker for the missing exhibit, reformatting the submission, navigating the rater, re-entering data the broker already entered. None of that is underwriting. All of it costs underwriting capacity.

The fastest way to scale underwriting is not to hire more underwriters. It is to give them back the 70%. At Atlas, the offshore BPO handling setup and clearance is exactly how 10–12 underwriters processed nearly 30,000 submissions a year. At PacificComp, a single underwriter managed a $32 million STP book at a 25% in-force loss ratio because the portal had absorbed every administrative task. The underwriter was free to make the call that only an underwriter can make. At BSB right now, the same logic is why portal entries have grown from under 50,000 a year to over 100,000 without the underwriting team scaling proportionally.

That separation of duties is the entire game.

Three things.

First, finding information that already exists somewhere in the company — policy details, prior loss runs, agent performance history. The data is there; it's just not where the person who needs it is looking. Second, re-keying data the broker already entered into a comparative rater into the underwriting system. Third, status-update theater — meetings, emails, manually built scorecards — that exists because the underlying system can't surface what's already true.

AI and modern infrastructure should be eating all three of these in the next 24 months. Most of them haven't yet. The carriers and MGAs that move on category one — AI that surfaces existing intel — will compound for two years while the laggards build category three software that automates a rate card.

This is a framework question. The right answer depends on where you are in your maturity curve and what kind of business you're processing. Here's the framework I use.

Tool one: portals. Portals are your foundation for standardized, repeatable business. At PacificComp the straight-through-processing portal — 100 eligible class codes (with the top 10 dialed in cleanly), up to $50K in premium, instant quote and bind — grew from $250K to $32M at a 25% in-force loss ratio, managed by a single underwriter. That's the power of a well-designed portal. It is practically self-operating for business that fits cleanly within your appetite — especially when paired with the right distribution. The cluster and aggregator partnerships — ISU Steadfast, Renaissance, Combined Agents, Pacific Interstate, United Agencies, We Insure — were the supercharger that drove submission flow into the portal. If it's standard and it's repeatable, portal it.

Tool two: BPO teams. BPO makes sense for volume tasks that require human processing but not senior underwriting judgment. At Atlas the offshore BPO team in India was transformational. They set up every account in Duck Creek overnight while we slept. 10–12 underwriters walked in each morning to ready-to-work queues. Nearly 30,000 submissions a year. $20M to $150M in two years. The BPO handled setup. The underwriters handled decisions. If it's high-volume, human-touch, but not a judgment call — BPO it.

Tool three: automation and workflow systems. These are for the connective tissue. At Kinetic, Novidea gave us real-time submission tracking, automated follow-ups, and cross-team collaboration between underwriting, claims, the wearable-tech team, and loss control. At PacificComp, Salesforce auto-emailed monthly scorecards to both branch leaders AND agency leaders the same day — and captured competitive intel on 90% of large quoted accounts. The newer agentic submission triage stack — Convr, Sixfold, Kalepa, Bold Penguin — is doing for 2026 what offshore BPO did for 2012: scoring every submission against guidelines, surfacing the clean fits, routing the rest. These systems don't replace people. They make people faster and more accountable.

So the framework. Portals for standardized self-service. BPO for high-volume human processing. Automation for routing, tracking, and accountability. One underwriter managed $32 million on a portal. Twelve underwriters managed $150 million with a BPO. Match the tool to the task.

Force the portal too hard and you lose submissions you would have written. Accept everything however it arrives and you lose underwriting capacity to formatting.

The honest answer is segmentation. Brokers placing small-business volume — agencies running 50+ accounts a month — want the portal because efficiency is their problem too. Brokers placing middle-market accounts at $250K+ premiums want the relationship and the dialogue. Trying to force the second group into a portal is a fast way to lose them.

At PacificComp we ran both lanes deliberately. Inside the same turnaround, two channels. The STP portal handled the small-business book — $250K to $32M with one underwriter, supercharged by cluster and aggregator distribution. The middle-market team — 14 underwriters working with Top 100 retail and specialty wholesale brokers — ran consultatively on accounts $250K+ and carried the bulk of the $65M to $311M book. Same carrier, two products, two channels, two design decisions. The producer's job is to know which broker fits which lane.

It makes sense when the process upstream of the BPO is well-defined, the data fields are stable, and the BPO is doing repeatable work the in-house team shouldn't be doing. It becomes a patch when the in-house process is broken, undocumented, or constantly changing — because then the BPO becomes the lowest-paid people doing the most ambiguous work.

We made BPO work at Atlas because we wrote the process down first. We knew exactly what 10–12 underwriters needed for clean intake in Duck Creek, and the BPO was scoped against that spec. The team in India operated overnight while we slept and delivered files ready to work in the morning.

If you're using BPO to avoid writing the spec, the BPO will fail and you will blame the BPO. The spec was the problem.

Outsourcing moves the cost. Improving the workflow removes the work. They look the same on the P&L for a quarter or two, then diverge sharply.

If you outsource a broken workflow, you've made the broken workflow cheaper to run — but you haven't fixed the underlying problem. And you've created a dependency on a vendor who profits from the brokenness staying broken.

The discipline I've tried to keep: spec the workflow first, simplify what can be simplified, automate what's truly repetitive, and only then decide what to outsource. Most teams reverse that order and wonder why their BPO contract keeps creeping.

Three things, and they're not technical.

First: a clear owner with the authority to make trade-offs and the political cover to make unpopular ones. Second: the work was scoped against actual broker and producer behavior, not against what the org chart wished broker behavior was. Third: the rollout was sequenced — one segment, one product, one geography — not a big-bang.

The Salesforce rollout we built at PacificComp — and extended across CA, NV, and AZ post-acquisition under CopperPoint — worked because we had executive air cover, we built it for how the BDMs actually worked rather than how we wished they worked, and we phased it. The auto-emailed monthly scorecards going to both branch leaders AND agency leaders broke the carrier-as-judge dynamic. The competitive intel captured on 90% of large quoted accounts fed directly to our Chief Actuary for rate studies by class code, premium band, and ZIP.

The rollouts that fail usually have all three issues at once — diffuse ownership, abstract scoping, and a launch date driven by the calendar instead of the readiness.

Timing is everything in technology adoption. I think about it through the lens of competitive advantage versus operational risk. Being early can be a massive differentiator. Being too early can be an expensive distraction.

Be early on technology that directly touches the broker experience. At PacificComp our small-business instant-bind portal went from $250K to $32M over six years at a 25% in-force loss ratio because we gave agents something nobody else was offering — speed. 100 eligible class codes (with the top 10 dialed in cleanly), up to $50K in manual premium, instant quote and bind. One underwriter managing the entire book. That wasn't just operational efficiency — that was a structural advantage in submission flow that drove growth even when competitors tried to match our pricing. When brokers love the experience, they send you more business. Period. WTW's March 2026 Advanced Analytics & AI Survey put hard numbers on this: insurers using sophisticated analytics achieved combined ratios six percentage points lower and premium growth three percentage points higher than slower adopters. That's not a marginal advantage. That's compounding.

Wait on technology that requires your internal processes to be mature first. At Kinetic we deliberately used Google Docs and email for our first 14 months — taking us all the way to $45M — because we needed to understand our workflow before designing the right system. Our technical team from the wearable side pushed Google's capabilities further than most people thought possible. And when we finally implemented Novidea, adoption was immediate — because we knew exactly what we needed. We had lived through the pain. The waiting wasn't laziness. It was intelligence.

The test I use, borrowed from Charlie Munger's inversion principle. Instead of asking "what will this technology do for us?" — flip it.

What happens to us if our competitors adopt this and we don't?

If the answer is "we lose distribution flow" or "we lose broker mindshare" — move now. If the answer is "we'll be slightly less efficient internally" — you have time.

So the framework. Be early on broker-facing technology. Be deliberate on internal operations technology. Use the inversion test to prioritize. That keeps you competitive without chasing every shiny object in the room.

They modernize the front end first — comparative rater, portal — without fixing the data and underwriting workflow underneath. Then they're shocked when the portal generates volume the underwriters can't process, and the team's broker relationships suffer because turnaround times collapse.

Or they modernize the CRM without changing the management discipline that uses it, and six months later the new CRM is just as half-populated as the old one.

Modernization is a behavior project disguised as a technology project. If you don't change how the team manages the work, the new tools just create a more expensive version of the old problems.

The companies that get it right have the discipline to do the operating-model work BEFORE the technology purchase. The ones that get it wrong assume the technology will install the discipline.

Critical, and underweighted.

When the producer is selling against a quote the underwriter doesn't see in real time, when the ops team is reporting numbers nobody else trusts, when each function has its own version of the truth — the friction is enormous and almost entirely invisible because everyone is too busy reconciling to notice.

The Salesforce scorecards we built at PacificComp — and extended under CopperPoint — went to both branch leaders AND agency leaders the same day, which broke the carrier-as-judge dynamic. Both parties saw the same numbers. That same principle scales: when producers, underwriters, and ops are looking at the same dashboard, the conversation moves from "whose data is right?" to "what should we do?" That shift alone is worth a year of growth.

The intel-capture loop matters here too. At PacificComp, capturing competitive intel on 90% of large quoted accounts and feeding it directly to the Chief Actuary for rate studies by class code, premium band, and ZIP meant the rate filings were informed by the field, not by the actuarial team's best guess. That feedback loop is what separates a carrier that prices reactively from one that prices proactively.

The first signal is the broker stops sending their best work. They don't tell you — they just route the cleaner accounts to the carrier who responds faster, and they leave you with the harder cases. By the time the leadership team notices, the loss ratio has been creeping for three quarters.

The second signal is producers who used to be your champions start sitting on the broker advisory committee for somebody else.

Operational quality is the most important part of the broker relationship that nobody includes in a relationship discussion. They talk about lunch and golf. They mean turnaround time, transparency, and intel.

There's a sales discipline I've leaned on across multiple builds: joint venture point-of-sale. At Pride Risk we sold complex payroll and workers' comp under a PEO agreement by joining the agent at the table — no one could sell our program better than we could. Closing ratio moved from 10% to over 20%. At PacificComp we ran the same play in the middle market on accounts $250K to over $1M. We brought PacificComp executives to the table alongside the broker — in-house claims and legal, industry vertical expertise, and hands-on loss control. Middle-market hit ratio went from below 15% to over 20%. When the carrier shows up at point-of-sale with real expertise — not just a quote — the broker remembers. And the next account is yours.

At PacificComp — coming out of a combined ratio north of 200% — rebuilding broker trust required operational discipline, not just relationship work. We instituted weekly large-account triage to prioritize wins and reasons to move. We secured 25% of large policy renewals — over $500K premiums — with commitments 60 days in advance. That wasn't a relationship trick. It was an operational discipline that made producers want to bring us the deal. The book went from $65M to $311M, returning to operating profit in 2016 and underwriting profit in 2017. Operations was the work that earned the relationships back.

This is the question every leader should be asking, because the answer unlocks leverage-based growth — more output per person without burning people out. Three levers, and I'm running all three right now at Berkley Small Business.

Lever one: separate intake from underwriting. This is the single biggest lever I've ever pulled. At Atlas, 10–12 underwriters handled nearly 30,000 submissions a year because the offshore BPO did all the setup work overnight. At PacificComp, one underwriter managed a $32 million STP portal book while 14 middle-market underwriters handled the rest of the $311M in-force book. Neither of those required hiring proportionally more underwriters as volume grew. You design the intake, and the leverage is built in. At BSB, that same principle is why we've gone from under 50,000 portal entries a year to over 100,000 without growing the underwriting team proportionally.

Lever two: distribution discipline — with teeth. The math: if you need 6,500–7,000 submissions a month to hit your growth targets at a 30% hit ratio, every single agency needs to earn its slot. At Atlas, we put the bottom 50 agencies on structured success plans and terminated 50+ who couldn't perform. At BSB right now, I refocused the wholesale trading-partner network from 42 to 24 — fewer, deeper, better fit. The capacity you recover from underperformers is as valuable as a new hire. You freed up underwriting capacity, field time, and management attention for your top tier — the agencies and trading partners actually driving the book. You don't need more headcount if your existing distribution is working harder and smarter.

Lever three: front-load information capture. At PacificComp, we added vertical-specific intake questions — auto dealers had their own set, hotels had theirs, grocers had theirs — and shut down the submission email box entirely. No exceptions. That single decision shrank the clearance team from five people to two and eliminated the endless back-and-forth emails killing turnaround time. Three full-time equivalents of capacity recovered through design, not hiring.

So the three highest-impact levers. Separate intake from underwriting. Implement real broker accountability to reallocate capacity. Front-load information capture to stop chasing what should have been provided upfront. Each one creates leverage. More output per person. That's how you grow smart, not just grow big.

I'm sitting in the middle of this answer right now. At Berkley Small Business I serve on the AI Committee, and AI shows up in my work every week — on real producers, real territories, real accounts. Not theory. Not someday. So when I tell you what the winners will do differently, I'm describing what we're already building.

Three things will define the platforms that win the next five years.

First, the best platforms will achieve what I call "invisible underwriting." Straight-through processing for clean risks. API connectivity to data sources and partners. The ability to change product rules and rating weekly or biweekly — not annually. At PacificComp a decade ago, our STP portal gave agents instant quote and bind — 100 class codes (with 10 dialed in cleanly), up to $50K in premium, $250K to $32M at a 25% in-force loss ratio, one underwriter. That was revolutionary then. The platforms that extend that speed to middle-market complexity — with AI-assisted prefill, automated triage, and real-time portfolio analytics — will fundamentally change broker expectations.

Second, the best platforms will build data moats through their distribution networks. Every submission, every quote, every win and loss is a data point. At PacificComp, we captured competitive intel on 90% of large quoted accounts and fed it directly to the Chief Actuary for rate studies by class code, premium band, and ZIP. The Salesforce scorecards went out the same day to both branch leaders and agency leaders — same numbers, no carrier-as-judge dynamic. McKinsey calls that "the data flywheel": better data leads to better pricing, which leads to better results, which attracts more brokers, which generates more data. The platforms that build that flywheel will be nearly impossible to displace.

Third, the winners will blend technology and human expertise in ways we haven't seen before. At Kinetic, wearable technology and AI claims tools gave us a value proposition no amount of pure underwriting skill could replicate. DHL — a Fortune 500 — was a self-insured client on the wearable side, and that was our proof point walking into every last-mile account. But it was the combination of that technology with experienced distribution professionals and consultative underwriters that took us from zero to $45 million in 18 months. Not the tech alone. Not the people alone. The combination.

Five years from now, the best MGA and MGU platforms will look like fintech companies in speed, data sophistication, and reach — but they'll still be grounded in underwriting discipline and broker relationships.

That's the future I'm building toward at Berkley Small Business, and it's the future Seven16 Group exists to serve — for the operators most insurance technology forgets.

Technology enables scale. Judgment and relationships remain the moat. That's the work.