Large insurance organizations carry more process complexity than most industries. Guarantees, compliance cycles, underwriting workflows, renewals, reserving, claims. The kind of work where a missed deadline does not appear in a dashboard. It appears in operations, weeks later, as an exposure somebody has to resolve.
Now distribute that across sixty operating units, each with its own systems, its own history and its own way of running the same nominal process. The group has a view of financial results and almost no view of how the work that produces them is actually performed.
This is the structural condition of specialty insurers, holding companies, acquisitive groups and federated organizations generally. It is not a reporting failure. It is what decentralization does, and it has specific consequences.
Why a group cannot see its own operations
Four mechanisms, and they compound with each acquisition.
Each unit reports upward in its own terms. A unit describes its process using its own names for stages, its own definition of a case and its own exception categories. Rolled up, those become a number with no comparable content behind it.
Autonomy was the reason for the structure. Operating units are frequently independent because independence is commercially useful: local underwriting judgment, specialist knowledge, speed of decision. Standardizing the description can be read as standardizing the operation, which is what the structure deliberately avoided.
System diversity makes central data pulls incomplete. Twenty policy administration systems means a central query reaches the common fields and nothing else. The common fields are the ones that were standardized precisely because they were not distinctive.
Processes with no system record are invisible at every level. The guarantee that expires, the compliance cycle that runs on a spreadsheet, the renewal that depends on somebody remembering. These leave no trace in any system, so they are absent from unit reporting and therefore absent from group reporting.
The specific risk in long-cycle work
Insurance operations have a property that makes this worse than it is in faster industries. Many obligations are time-bound and far out: a guarantee with a multi-year term, a compliance filing on an annual cycle, a renewal window, a reserving review.
A missed step in a fast process surfaces within days because something downstream stops. A missed step in a long-cycle process surfaces when the deadline arrives, which can be quarters later, and by then the options have narrowed considerably.
| Process type | Time to detect a missed step | Where it shows up |
|---|---|---|
| High-volume transactional | Days | Queue backlog, service metrics |
| Claims handling | Weeks | Cycle time, complaints |
| Underwriting exceptions | Weeks to months | Loss ratio, audit findings |
| Compliance cycles | At the filing deadline | Regulatory exposure |
| Guarantees and long-term obligations | At expiry or call | Direct financial exposure |
The bottom two rows are where the cost concentrates and they are the rows with no early signal. A dashboard built on transactional volume will look healthy throughout.
Sixty units
What the group can see at each layer
| Layer | Visible centrally | Comparable across units |
|---|---|---|
| Financial results | Yes | Yes |
| Volumes and headcount | Yes | Mostly |
| Systems in use | Yes | Not meaningful |
| Process stages as each unit names them | Partly | No |
| How the work is actually performed | No | No |
| Steps with no system record | No | No |
The top two rows are what group reporting contains. The bottom three are where long-cycle risk lives, and none of them is comparable without a deliberate effort to make it so.
What standardization should and should not touch
The useful distinction in a federated group is between the description of the work and the performance of it. A group can standardize how units describe their processes without standardizing how they run them, and that is the move that makes oversight possible without removing the autonomy the structure exists to protect.
| Element | Standardize across units | Leave to the unit |
|---|---|---|
| Definition of a case or policy unit | Yes | No |
| Names of process stages | Yes | No |
| Exception categories | Yes | No |
| How hours and effort are recorded | Yes | No |
| Underwriting judgment and appetite | No | Yes |
| Local regulatory handling | No | Yes |
| Systems and tooling | No | Yes |
| Sequence and local adaptations | No | Yes, with the reason recorded |
The first four rows cost a unit almost nothing and they are what turns sixty reports into one comparable picture. The last four are where the commercial value of independence sits.
Governance standards describe this split directly. ISO 37000, which gives guidance for the governance of organizations, treats oversight as requiring information that the governing body can actually use, distinct from management's own reporting. ISO 9001 requires an organization to determine its processes and their interactions and to control processes performed by external parties on its behalf, which is the position a group holds relative to an operating unit it does not run day to day.
The practical problem with the usual approaches
| Approach | What it produces | Limitation in a multi-entity group |
|---|---|---|
| Central data warehouse | Comparable financial and volume data | Reaches common fields only, misses unrecorded work |
| Unit self-assessment questionnaires | Each unit's own account | Not comparable, and reported upward with incentive |
| Internal audit cycle | Deep view of a few units per year | Coverage of sixty units takes many years |
| Traditional consulting review | A thorough account of selected units | Cost per unit makes full coverage impractical |
| Observation tooling | Activity on screens where deployed | Deployment across twenty system estates is its own programme |
| Interviews at census scale per unit | Comparable descriptions from the people doing the work | Requires the same instrument run in every unit |
The constraint running through the middle rows is cost per unit. Any approach priced per engagement makes full coverage of sixty units a multi-year programme, and by the time it finishes the early units have drifted.
What a usable group-level view contains
Four things, and the fourth is the one that is usually missing.
The same process described in the same terms by every unit. Not the same process. The same description standard, so a renewal in unit twelve and a renewal in unit forty can be read against each other.
Hours and effort attributed to named activities. A weekly figure on a named activity in a named unit, which is comparable and claimable. A percentage of capacity is neither.
The exception inventory per unit. What departs from the normal path, how often, and what happens then. This is the content that predicts a missed long-cycle obligation and it exists nowhere in system data.
Which steps have no system record. The list of activities running on spreadsheets, email and memory. This is the group's actual risk register for operational surprise, and almost no group has one.
Where Horizon fits
Horizon is an AI-powered continuous discovery platform, and the multi-entity problem is largely a cost-per-unit problem. A description of how work runs in one unit is useful. The same description in sixty units, produced to a comparable standard, is a different kind of asset, and it only exists if producing each one is cheap.
Discovery Cycles run AI-led interviews asynchronously with the people performing the work, adapted by role, with the same instrument in every unit. That is what makes the outputs comparable: the questions do not change because the unit changed, so a renewal process in one entity can be read against the same process in another without a translation step.
The Process Library holds each unit's described process, including the steps that leave no system record, which is where the long-cycle exposure sits. The Insights Dashboard attaches hours and impact to findings with traceability to the inputs behind them, so a group-level figure can be opened down to the specific unit and activity. Workspaces keep each entity separated with anonymization, which matters in a federated structure where a unit's willingness to be candid depends on knowing what the group sees.
At La Segunda, a specialty insurer, Horizon worked with a nine-person team and covered 100% of it, producing 10 findings against roughly 700 active cases and 48 hours of process mapping. Participants rated the experience 8.2 out of 10 and 90% said they would have the conversation again. A nine-person specialist team is the unit-level version of this problem, and census coverage at that size is what makes the unit comparable to the next one rather than sampled. These figures describe one engagement and they are evidence rather than a guarantee.
The question worth asking at group level: for each operating unit, do we have a list of the activities that leave no system record, and was it produced to the same standard everywhere?
Common mistakes
Treating the problem as a data integration project. A central warehouse reaches the fields the units have in common, which are the fields that were standardized because they were not distinctive.
Standardizing the operation instead of the description. The autonomy was commercially deliberate. The description standard is what oversight needs, and it costs a unit almost nothing.
Relying on unit self-assessment. Each unit reports in its own terms and with its own incentive, which produces sixty accounts and no comparison.
Building oversight on transactional volume. The risk concentrates in long-cycle obligations, which produce no volume signal until the deadline.
Rolling out audit coverage sequentially. At a few units per year, the early units have drifted before the last ones are reached.
Omitting the unrecorded-work inventory. The list of activities running on spreadsheets and memory is the group's real exposure list, and most groups do not have one.
Letting each unit choose its own exception categories. It is the cheapest thing to standardize and the one that makes comparison possible.
FAQ
Why can't a group see how its operating units actually work?
Because each unit reports upward in its own terms, system diversity limits central data pulls to common fields, and any activity that leaves no system record is invisible at unit level and therefore at group level. Financial results roll up cleanly and operational practice does not.
What makes insurance operations harder than other industries here?
The length of the cycles. A missed step in a transactional process surfaces within days because something downstream stops. A missed step in a guarantee, a compliance filing or a renewal surfaces at the deadline, which can be quarters later, when the options have narrowed.
Does standardizing processes across units remove the value of decentralization?
Standardizing the description does not. A group can require the same definition of a case, the same stage names, the same exception categories and the same way of recording effort, while leaving underwriting judgment, local regulatory handling, systems and sequence to the unit. That split preserves the autonomy and produces the comparison.
Why does internal audit not solve this?
Coverage rate. An audit cycle gives a deep view of a few units per year, so full coverage of several dozen units takes many years, and the units examined first have changed before the last ones are reached.
What should a group-level operational view contain?
The same process described in the same terms by every unit, hours attributed to named activities, an exception inventory per unit, and a list of the activities that leave no system record. The last one is the closest thing a group has to an operational risk register and it is usually absent.
Comparable descriptions are the oversight asset
A federated group does not lack data. It lacks sixty accounts of the same process written to one standard, which is a different thing and the only version that supports a decision at group level.
The content that matters most is the part no system holds: what departs from the normal path, how often, and which obligations are being tracked by somebody remembering. Produced to the same standard in every unit, that becomes oversight. Produced once per unit in each unit's own terms, it stays sixty reports.
See it. Fix it. Lead it.