The Stack Nobody Chose: Best-of-Breed vs. Integrated Platform, With the Real Cost Math

Open the software tab of any mid-size MGA and you will find an archaeology site. The agency management system from eight years ago. A quoting tool bought for one program. A commission spreadsheet that is load-bearing infrastructure. A producer portal from a vendor who was acquired twice. E-signature documents living in one cloud, endorsements in another, and the bordereau, sacred and terrible, built fresh every month by whoever has the evening free.

Nobody in the room ever voted for fragmentation. Each piece was the reasonable choice at the moment it was made: the AMS solved the policy problem, the portal solved the distribution problem, the spreadsheet solved the problem in front of the person who bought it. Best-of-breed is what a series of locally rational decisions looks like from ten thousand feet. The bill for the collection, however, does not arrive itemized. It arrives as headcount, reconciliation time, and incidents no vendor's support ticket will ever own.

Where Best-of-Breed Genuinely Wins

Intellectual honesty starts with the real advantages, because they are not imaginary:

• Niche depth. A tool built for one job often does that job better than a platform's module does it. Standalone comparative raters, survey engines, and compliance calendars have historically out-classed suite features, because the suite splits its roadmap ten ways.
• No single point of failure. When one vendor's platform has an outage or ships a bad release, everything stops. A fragmented stack fails in pieces, and pieces are survivable.
• Switching leverage. Replacing one bolt-on is cheap. Replacing a platform is a project measured in quarters. Fragmentation keeps vendors honest on price and features.
• Speed of first use. Buying a point solution takes a week and a credit card. Platform decisions take committees. For a firm that needs something fixed by Friday, the stack is often the right tactical answer.

The Integration Tax: What Fragmentation Actually Costs

Here is where the invoice shows up. None of these line items carries a vendor's name, which is why they stay invisible in budget season:

1. Reconciliation headcount. Two systems of record means a human arbitrating them. In our work on back-office bottlenecks and commission accounting, the pattern repeats: the commission dispute is rarely about the commission rate. It is about the AMS, the billing system, and the spreadsheet disagreeing about premium, and a smart person spending three days deciding which one is right.
2. Data decay at every hop. Sync integrations preserve data at the moment they are written. Everything after that is drift. Producer changed in one system. Entity name corrected in another. The bordereau quietly assembled from the stale copy, because every copy in the stack is somebody's source of truth.
3. Integration maintenance, forever. APIs version, deprecate, and break. Each connector is a small plant that must be watered for as long as both vendors exist. Five tools means ten possible seams, and someone in the firm owns those seams as an unpaid side job, which is how operations quietly become fragile.
4. The audit surface multiplies. Under delegated authority, the evidence trail for one transaction can span four systems, each exporting differently. Reconstructing why a risk was bound, quoted, or endorsed turns into a forensics project every single audit cycle, as we covered in our piece on delegated authority and E&O exposure.
5. Analytics stay stuck at summary level. Nobody builds a cross-system data warehouse for a 40-person MGA, so the questions that need premium, losses, producer activity, and expenses in one join get answered by exports, pivots, and vibes. The competitive decisions your carrier makes about your program are increasingly informed by its own joined data. Yours are informed by whoever reconciles fastest.

Where the Integrated Platform Case Is Strong

The platform argument is not that one vendor is cleverer at every function. It is architectural: when submission, quoting, issuance, endorsements, claims, commissions, and carrier reporting read and write the same records, an entire category of work simply disappears. There is nothing to reconcile, because there is only one thing. The workflow runs end to end without a human acting as the API between systems, and the audit trail is a byproduct of doing the work instead of a reconstruction project. That is the difference between our loss run analysis feeding a live underwriting decision and a parsed spreadsheet feeding a copy-paste into a third system. And it is why integration capabilities matter most as an architectural property of one system, not as count of prebuilt connectors.

Platforms have honest weaknesses too: you bet your operations on one vendor's roadmap, you pay for modules you do not use, and the long-tail edge case that your niche program exists to write is the case their configuration cannot express. That last one, the 80 percent that is common and the 20 percent that is your actual business, is where the platform-vs-custom trade-off we mapped in build vs. buy becomes relevant.

A Decision Framework, Not a Conclusion

Firms usually make this choice with a spreadsheet of license costs, which is comparing the visible part of the bill. Try these five questions instead, and answer with names and hours:

• How many records cross system seams every week, and who, specifically, carries them?
• When the commission, the bordereau, and the GL disagree, how many days does it take to decide who is right, and who makes that call?
• What did integration maintenance cost last year, counting contractor invoices and the salary fraction of the person who quietly owns the connectors?
• Can you produce one complete transaction file, submission to claim to commission, as a single linked record in ten minutes? (The answer is a meeting, not a query.)
• Which tool would you fight to keep if you had to consolidate to two systems tomorrow? The strength of that feeling is the real best-of-breed argument, and its weakness is the integration tax you are paying to preserve it.

The pattern we see in firms that get this right is neither extreme: an integrated core that holds the policy lifecycle, the money, and the audit trail in one place, with specialized tools attached at the edges where a niche genuinely beats a suite. The core never asks a human to be the integration. The edges are replaceable by design. That middle position is an architecture decision, which is why we treat it as consulting work: a technology audit of the seams first, then the build, on platforms like InsuranceClouds when an integrated core is what the math supports, or through our development team when the 20 percent is the business.