Foundry Operating Thesis
Operating Model
Foundry operates as a holding company with sector-focused operating platforms. Individual operating platforms focus on clusters of businesses with similar workflow structures - this structure allows each platform to develop domain expertise while leveraging shared transformation infrastructure developed centrally.
Financial Workflow Platform
Our initial area of focus is a class of financial services businesses that operate structured, compliance-driven workflows. These companies sit within the operational infrastructure layer of the financial system, enabling financial risk and obligations to be evaluated, documented, executed and monitored.
Across insurance, lending, servicing and compliance, the same underlying architecture recurs: document-heavy intake, rules-based decisioning, structured case management, regulatory verification and repeatable reporting. These shared primitives are more important than the end-market category and define the true boundaries of the sector.
This common operating model makes the platform particularly well suited to AI-enabled transformation. High levels of manual processing, fragmented data and human-in-the-loop decisioning create clear opportunities for standardisation, automation and margin expansion.
Insurance workflows remain the most attractive initial entry point due to their density of these primitives, but the broader Financial Workflow Platform provides a more scalable framing than a narrow MGA-only lens. We start with MGA because it provides a practical environment in which to deploy the operating system, establish proof of performance, and build expertise that can later expand into adjacent insurance and financial workflow categories.
Acquisition Strategy
We’re building an operating system for AI transformation and deploying it into increasingly bigger deals, laddering up towards the enterprise.
Larger businesses contain more workflow volume, more data, and greater embedded inefficiency. As a result, the same system produces disproportionately greater gains when deployed at scale. Each successive acquisition increases both the size of the platform and the effectiveness of the system applied to it.
The system improves with each deployment, becoming more capable, more reusable, and faster to implement. This creates a compounding effect and the result is a non-linear scaling dynamic. Within a small number of transactions, the holdco can transition from initial deployments into enterprise-scale deals.
Workflow primitives and business archetypes
The most useful way to analyze this market is through workflow primitives. These are the repeatable operating units that recur across very different-looking businesses: intake, extraction, verification, routing, decision support, servicing, settlement and reporting. The same primitives reappear across delegated underwriting, claims administration, specialty lending and compliance operations. That is what creates the possibility of a shared automation stack.
- Delegated underwriting / MGA businesses concentrate risk evaluation, pricing, broker coordination, policy issuance and bordereaux production.
- Insurance services, claims businesses, and carriers concentrate case management, evidence handling, QA, customer communication, reporting, and core underwriting workflows.
- Specialty lenders and servicers concentrate document intake, credit verification, collateral review, servicing and collections.
| Archetype | Role in ecosystem | Common primitives | Typical monetization |
|---|---|---|---|
| MGA | Underwrites and distributes on behalf of carriers | Intake, risk evaluation, routing, pricing, issuance, bordereaux | Commission, fees, profit share |
| Insurance services / claims | Administers claims, policies and operational support | Intake, case management, QA, compliance, reporting | Admin fees, service contracts |
| Specialty lending / servicing | Underwrites niche credit and services loans | Intake, credit verification, collateral review, servicing, collections | Interest margin, origination, servicing fees |
Financial Workflow Market Map
Our initial market map shows the distribution of companies in the UK with workflows most closely overlapping with our starting model, MGAs. The objective is to map the space and identify patterns, not to present diligence-grade financials for every company.
Estimated EV range
£10m to £1bn.
Mapped set
469 companies across 11 sectors and 4 broad archetypes.
Market shape
63% of mapped names sit below £50m estimated EV, 87% sit below £100m, and 10 companies sit at or above £250m EV.
This is attractive for sourcing our first deals and proving expertise in transforming companies within less-regulated insurance, and specialty lending. Ultimately, the broader insurance and financial workflows space provides almost unlimited scale within the UK alone, let alone other geographies.
MGA Market Map
Our current UK mapping identifies circa 180 MGA businesses, with a strong skew toward the lower mid-market and a number of scaled platforms. This distribution makes MGA an accessible starting point.
UK MGA enterprise value distribution
Estimated enterprise values are grouped into fixed EV bands to show where fragmentation and scale sit inside the mapped set.
Hover a bar to isolate an EV band within the mapped set.
The market is sufficiently fragmented to provide entry opportunities, whilst still offering enough depth in the small and mid-market to support subsequent acquisitions and early platform development. It allows us to deploy the operating system in live environments where workflows are clear, measurable, and economically meaningful.
At the same time, the supply of large, enterprise-scale assets within MGA is limited. We see meaningful opportunity to compete organically at the upper end of the market.
That said, MGA is not the destination of the strategy but rather the entry point. It provides an ideal environment in which to establish the system and generate early proof of performance. From there, expansion follows into adjacent parts of the insurance value chain and into the broader Financial Workflow Platform, where similar primitives are found as well as businesses at significantly greater scale.
MGA sector deep dive
We're starting with MGAs because they are durable, embedded, and operationally intense.
MGAs are deeply integrated into the insurance ecosystem. They perform a core economic function, underwriting and distributing risk, within a regulated framework that is unlikely to be disintermediated. Demand is persistent, and their role in connecting brokers, carriers, and insured parties is structurally embedded.
This satisfies the first part of our lens: we are acquiring into a category that should still exist over the long term. Within that durable category, MGAs are highly operational businesses. Their performance is not primarily constrained by demand or pricing power, but by how effectively work is executed. Core functions such as submission handling, underwriting support, policy administration, and reporting remain heavily manual, often relying on fragmented systems and repeated human intervention.
As a result, a significant portion of cost is concentrated within a relatively small set of repeatable workflows. Improving how this work is executed does not just reduce cost, it enhances the competitive position of the business.
More efficient underwriting workflows improve responsiveness and consistency, strengthening broker relationships. Better data capture and processing improve risk selection and reporting, strengthening carrier relationships. Reduced operational friction allows higher-quality underwriters to focus on decision-making rather than process, improving talent density and retention.
In this way, productivity gains compound beyond margin expansion. They reinforce the core relationships and capabilities that define long-term success in the market.
Ideal MGA targets
We seek MGAs with embedded franchise value and clear upside from automation.
Embedded value comes from access to capacity from multiple carriers; durable broker distribution; and deep expertise in lines where conditions are likely to harden over time.
In order for automation to materially uplift EBITDA, MGAs must have labour-heavy opex, a high proportion of manual workflows, high spend on software and IT consultancy. Where automation matters most is not necessarily in pricing. It is in the heavy, repetitive operating flows around e.g. submissions, referrals, bordereaux, endorsements, renewals. Addressing these these properly can create value without needing a heroic growth case.
They should be specialist enough to maintain margins, sufficiently high volume to deliver returns from automation, and big enough that carriers and brokers care about service quality.
We like SME and lower mid-market commercial lines (eg cyber for SMEs, professional indemnity, and contractor/trades liability), where submission volume is high and workflows are messy. Specialist commercial niches such as marine cargo, fleet/commercial auto, and renewable energy contractors packages are also attractive, because they have real underwriting IP but enough volume to industrialise parts of the workflow.
| Screening Lens | What matters |
|---|---|
| Franchise durability | Multi-carrier capacity, broad broker access, respected underwriters, harder markets over time |
| Underwriting quality | Strong market standing, real niche expertise, trusted by brokers and carriers |
| Labour intensity | Large share of opex in people and support functions |
| Workflow intensity | Heavy manual work across submissions, endorsements, renewals, bordereaux, reporting etc |
| Revenue upside | Faster, more accurate service can win more broker flow |
| Carrier upside | Better reporting, controls and execution can deepen capacity access |
| Talent upside | Better systems let underwriters spend less time on admin and more on judgement |
The UK MGA market is attractive because it combines capital-light underwriting models with genuine specialization, fragmented asset supply and a growing delegated authority backdrop.
Value creation levers
| Value creation lever | Atomic task breakdown | Inputs | Baseline cost | Cost post-automation | Net arbitrage |
|---|---|---|---|---|---|
| 1. Submission Intake, Data Capture & File Structuring Approach to automation: The real labour pool is not opening emails. It is turning incomplete broker packs into decisionable files before an underwriter ever touches them. Faster intake improves broker responsiveness and keeps underwriting time focused on risk, not reconstruction. | 1.1 Ingestion: Capture submissions from email, portal, and attachments into one case record. | Broker email or portal submission, slip, schedule, prior terms. | £3 (~5m Ops) | £1 Rationale: Automated inbox and portal capture creates the case instantly; humans only step in when files are malformed. | £2 |
| 1.2 Extraction: Parse slips, schedules, and attachments into structured quote-ready fields. | PDF slip, Excel schedule, ACORD forms, supporting attachments. | £11 (~16m Ops) | £2 Rationale: Extraction models turn unstructured pack data into JSON, with QA reserved for noisy documents and awkward schedules. | £9 | |
| 1.3 Data Entry: Write extracted data into the PAS or structured case record so the file is operable downstream. | Structured risk fields, schedules, and PAS or workflow schema. | £6 (~9m Ops) | £1 Rationale: Direct API push and structured forms remove most rekeying while still allowing human QA on messy schedules. | £5 | |
| 1.4 Completeness chase: Flag missing data and issue broker questions to make the file decisionable. | Structured submission, appetite checklist, and missing-field logic. | £6 (~6m Ops, 2m Comm) | £1 Rationale: Systems can generate targeted chase packs, but people still handle ambiguous exposures and broker negotiation. | £5 | |
| 1.5 Triage: Route the case to the right product, authority, and underwriter queue. | Structured submission, product taxonomy, and routing rules. | £3 (~4m Ops) | £1 Rationale: Workflow routing can be automated once the case is structured, leaving only true exceptions for manual allocation. | £2 | |
| Subtotals (annualised) | Volume: 40,000 submissions /yr | £1,160,000 | £240,000 | £920,000 | |
| 2. Underwriting Review, Referral & Quote Rework Approach to automation: This is the judgement-heavy centre of the MGA. The gain is not replacing underwriters; it is stripping out rekeying, manual rule checking, and repetitive quote rework so scarce underwriters spend time on selection, pricing, and broker negotiation. | 2.1 External enrichment: Pull third-party data and assemble a decision-ready risk profile. | Structured submission, third-party data, prior loss history, sanctions data. | £14 (~8m UW, 4m Ops) | £4 Rationale: API enrichment and synthesis remove manual searching while keeping human review for unusual risks. | £10 |
| 2.2 Appetite check: Test whether the risk fits the target product and broker appetite for the MGA. | Risk summary, product rules, and appetite checklist. | £11 (~6m UW, 2m Lead) | £2 Rationale: Rules engines and model-assisted comparisons catch most breaches early; humans resolve the grey areas. | £9 | |
| 2.3 Authority check: Test line size, terms, and delegated authority limits against the intended quote. | Risk summary, authority matrix, and carrier rules. | £6 (~3m UW, 1m Lead) | £1 Rationale: Automated cross-checking catches authority breaches before they create costly referral loops. | £5 | |
| 2.4 Wording check: Confirm that the intended wording and exclusions align with product and binder rules. | Risk summary, binder wording, and wording constraints. | £6 (~3m UW, 1m Lead) | £1 Rationale: Model-assisted wording comparison removes routine checking while preserving human control over tricky edge cases. | £5 | |
| 2.5 Disposition: Auto-decline clear out-of-appetite risks, auto-approve clear in-appetite risks, and route the remainder into human review. | Decisioned risk profile, authority rules, and hard-stop logic. | £9 (~6m UW) | £1 Rationale: Automated disposition creates team leverage by reserving underwriter time for the genuinely complex middle of the distribution. | £8 | |
| 2.6 Referral pack preparation: Prepare the escalation pack for senior underwriters or carrier sign-off. | Risk profile, authority breach notes, and referral precedents. | £14 (~6m UW, 4m Lead) | £3 Rationale: Systems can assemble the facts, but humans still frame the actual commercial and authority judgement for referral. | £11 | |
| 2.7 Rating: Apply rate manual multipliers and pricing factors to the validated risk profile. | Enriched risk file, rate manual, and pricing factors. | £29 (~20m UW) | £12 Rationale: Algorithmic rating removes repeat calculation work while leaving humans to review the difficult edge cases. | £17 | |
| 2.8 Endorsements: Select required exclusions and draft manuscript endorsements based on exposure. | Rated risk profile, wording library, and endorsement templates. | £20 (~14m UW) | £8 Rationale: Model-assisted wording selection removes routine drafting while keeping human review on bespoke terms. | £12 | |
| 2.9 Quote assembly: Generate the formal quote pack and broker-facing documentation. | Final terms, pricing, wording selections, and document templates. | £3 (~4m Ops) | £1 Rationale: Document assembly is mostly mechanical once terms are finalised. | £2 | |
| 2.10 Quote rework and repricing: Update terms after broker pushback, negotiation, or carrier feedback. | Quote pack, broker feedback, and carrier comments. | £12 (~6m UW, 4m Comm) | £5 Rationale: Automation accelerates comparison and redrafting, but the trading decision still sits with humans. | £7 | |
| Subtotals (annualised) | Volume: 15,000 quoted or referred cases /yr | £1,860,000 | £570,000 | £1,290,000 | |
| 3. Bind, Issuance & Final Deal Capture Approach to automation: A surprising amount of downstream pain starts here. If the MGA fails to capture what was actually agreed at bind, service, finance, and bordereaux teams inherit the clean-up. | 3.1 Final terms capture: Record the final negotiated version of limits, premiums, commissions, and endorsements. | Final quote, broker acceptance, email amendments, binder terms. | £8 (~5m Ops, 3m UW) | £1 Rationale: Structured bind workflows reduce rekeying, but humans still confirm the awkward last-minute changes that move the economics. | £7 |
| 3.2 Binder-to-PAS reconciliation: Confirm that the operational record matches the authority and the deal actually agreed. | Bound terms, PAS record, binder rules, and authority controls. | £6 (~4m Ops, 2m UW) | £1 Rationale: Automated comparisons catch drift before the wrong policy record flows downstream. | £5 | |
| 3.3 Compliance prep: Assemble sanctions, conduct, and compliance evidence linked to the bound deal. | Bound quote terms, sanctions data, and regulatory checklists. | £3 (~5m Ops) | £1 Rationale: AI-generated evidence packs remove routine compilation work while preserving human spot-checks. | £2 | |
| 3.4 Policy issuance: Generate policy documents, schedules, and evidence of cover on the right carrier paper. | Bound terms, document templates, carrier paper, tax rules. | £5 (~7m Ops) | £1 Rationale: Document generation and routing are largely mechanical once the final deal is captured correctly. | £4 | |
| 3.5 Broker dispatch: Deliver the bound documentation and evidence of cover to the broker cleanly and on time. | Issued policy pack, broker contact data, and dispatch workflow. | £2 (~3m Ops) | £0 Rationale: Automated dispatch removes repetitive sending and filing work. | £2 | |
| 3.6 Finance handoff: Create the premium, commission, and settlement record finance needs to operate. | Bound policy record, commission split, settlement references. | £4 (~4m Ops, 2m FinOps) | £1 Rationale: Straight-through data handoff removes manual re-entry and prevents later exception chasing. | £3 | |
| 3.7 Service handoff: Open the servicing record, diary tasks, and issue any in-force flags for downstream teams. | Bound policy record, servicing flags, and diary rules. | £3 (~5m Ops) | £1 Rationale: Workflow-triggered service handoff prevents later queue confusion and repeat data entry. | £2 | |
| Subtotals (annualised) | Volume: 10,000 bound policies /yr | £310,000 | £60,000 | £250,000 | |
| 4. In-force Servicing, Endorsements & Payment Friction Approach to automation: This is where mediocre MGAs quietly burn labour. Simple admin, awkward MTAs, cancellations, reinstatements, and collections issues sit in the same shared queue, so interruptions compound quickly. | 4.1 Service triage: Classify incoming queries and route them into the right service lane. | Broker email or call note, policy record, service request. | £3 (~5m Ops) | £1 Rationale: Automated triage and case creation reduce queue noise, but people still manage unusual requests and relationship-sensitive issues. | £2 |
| 4.2 Standard query response: Resolve routine broker and policyholder queries and create the right case record. | Service request, policy record, and response templates. | £4 (~6m Ops) | £1 Rationale: Template-driven responses remove repetitive drafting while keeping humans on relationship-sensitive service moments. | £3 | |
| 4.3 MTA assessment: Determine whether the requested change is straight-through or judgement-heavy. | Change request, existing policy, and change rules. | £5 (~3m Ops, 2m UW) | £1 Rationale: Rules-based classification ensures only the meaningful changes pull underwriters into the loop. | £4 | |
| 4.4 Re-rating: Recalculate premium and terms when the exposure change is economically material. | Existing policy, change request, and rating logic. | £8 (~5m Ops, 3m UW) | £2 Rationale: Straightforward recalculation can be automated, while underwriters focus only on changes that genuinely alter risk. | £6 | |
| 4.5 Endorsement drafting and reissue: Draft or select the right endorsement language and reissue the documents. | Approved change, wording library, and document templates. | £8 (~5m Ops, 3m UW) | £2 Rationale: Model-assisted drafting removes routine wording work while preserving human review on bespoke endorsements. | £6 | |
| 4.6 Cancellation handling: Issue notices, validate grounds for cancellation, and update the operational record. | Payment status, policy record, cancellation rules, and notice templates. | £4 (~3m FinOps, 2m Ops) | £2 Rationale: Workflow-triggered notices eliminate routine admin while keeping humans on disputed or commercially sensitive cases. | £2 | |
| 4.7 Reinstatement handling: Assess reinstatement requests and reopen cover where appropriate. | Reinstatement request, payment status, policy record, and approval rules. | £4 (~4m FinOps, 2m Ops) | £2 Rationale: Automated checks reduce repetitive validation work, but people still handle exceptions and judgement calls. | £2 | |
| 4.8 Collections chase: Chase brokers and counterparties for overdue premium and unresolved payment items. | Payment status, finance record, broker instructions, and chase rules. | £5 (~4m FinOps, 3m Ops) | £2 Rationale: Automated reminder logic shrinks the chase burden, but disputed or commercially sensitive cases still need people. | £3 | |
| Subtotals (annualised) | Volume: 18,000 service events /yr | £738,000 | £234,000 | £504,000 | |
| 5. Renewal Prep, Repricing & Retention Management Approach to automation: Renewals are not clerical. The best MGAs separate unchanged business from accounts that deserve real trading effort, get terms back early, and protect the base without over-renewing poor risks. | 5.1 Renewal diarying: Trigger the renewal workflow early enough to avoid last-minute firefighting. | Expiring policy list, renewal dates, and diary rules. | £4 (~6m Ops) | £1 Rationale: Automated diaries stop renewals from starting late and reduce manual queue management. | £3 |
| 5.2 Exposure refresh: Collect updated exposure and underwriting data from the broker before pricing begins. | Expiring policy, broker updates, and exposure schedules. | £5 (~4m Ops, 2m Comm) | £1 Rationale: Automated diaries and chase workflows prevent late surprises, with humans focusing on broker responsiveness and missing exposures. | £4 | |
| 5.3 Loss-run chase: Pull loss history and claims movement into the renewal file. | Expiring policy, loss runs, claims movement, and broker follow-up. | £3 (~2m Ops, 2m Comm) | £1 Rationale: Automated requests and ingestion remove repetitive chasing while still allowing humans to resolve missing or inconsistent claims data. | £2 | |
| 5.4 Pre-renewal technical review: Check the expiring risk against underwriting and conduct requirements before repricing. | Renewal file, claims movement, and underwriting guidelines. | £11 (~8m UW) | £1 Rationale: Summarisation and structured review reduce file-reading time while keeping humans in charge of risk judgement. | £10 | |
| 5.5 Re-rating: Apply updated carrier rates and pricing logic to the renewal risk. | Renewal file, updated rate tables, and pricing logic. | £23 (~16m UW) | £2 Rationale: Systems can re-rate and flag drift quickly; human effort stays on cases where pricing or appetite really changed. | £21 | |
| 5.6 Term comparison: Compare expiring versus new terms and make the economic changes visible immediately. | Expiring terms, repriced terms, and carrier guidance. | £9 (~4m UW, 2m Lead) | £1 Rationale: Automated comparisons make changes visible immediately, leaving humans to decide how hard to push. | £8 | |
| 5.7 Authority check: Confirm that the renewal terms still sit inside authority and carrier guidance. | Repriced renewal terms, authority rules, and carrier guidance. | £6 (~2m UW, 2m Lead) | £1 Rationale: Automated authority controls reduce manual rechecking and surface only the renewals that genuinely need escalation. | £5 | |
| 5.8 Renewal quote assembly: Generate renewal notices, quotes, or non-renewal documentation. | Final renewal terms, document templates, and broker contact data. | £4 (~6m Ops) | £1 Rationale: Document generation is mostly mechanical once the renewal decision is made. | £3 | |
| 5.9 Broker negotiation and retention decision: Negotiate the renewal and decide where to hold price, flex, or walk away. | Repriced terms, broker leverage, account priority, market intelligence. | £17 (~8m UW, 6m Comm) | £2 Rationale: Automation improves speed and preparation, but retention is still a trading decision rather than a pure admin workflow. | £15 | |
| Subtotals (annualised) | Volume: 10,000 renewals /yr | £820,000 | £110,000 | £710,000 | |
| 6. Premium Accounting, Bordereaux & Close Control Approach to automation: This is the hidden labour pool many investors miss. Dirty source data becomes undeniable here. A well-run MGA does not rely on heroic finance staff; it relies on clean identifiers, disciplined handoffs, and fewer exceptions to chase. | 6.1 Cash allocation: Match receipts to policies, statements, and brokers. | Bank receipts, broker statements, policy references, commission data. | £7 (~8m FinOps, 2m Ops) | £1 Rationale: Matching logic removes most routine allocation work, leaving humans to resolve genuinely ambiguous cash. | £6 |
| 6.2 Unmatched receipt chase: Investigate and chase cash items that fail to reconcile automatically. | Exception list, broker statements, and unmatched cash items. | £6 (~6m FinOps, 2m Ops) | £1 Rationale: Automated exception lists focus human effort only on the genuinely bad cash items. | £5 | |
| 6.3 Carrier settlement calculation: Calculate what is due to carriers and counterparties from the recorded business. | Carrier statements, cash ledger, premium movements, and settlement schedule. | £8 (~10m FinOps) | £2 Rationale: Structured reconciliation and evidence packs reduce manual close work, but humans still own material exceptions and carrier disputes. | £6 | |
| 6.4 Commission reconciliation: Reconcile commissions payable, receivable, and adjustments across the ledger. | Commission rules, policy records, broker statements, and settlement files. | £6 (~8m FinOps) | £1 Rationale: Structured matching removes routine spreadsheet work and surfaces only true anomalies. | £5 | |
| 6.5 Bordereaux build: Assemble bordereaux in the right carrier format from bound and changed business. | Bound and changed policy data, premium movement, and carrier templates. | £9 (~8m FinOps, 4m Ops) | £1 Rationale: Automated bordereaux generation removes formatting and collation labour. | £8 | |
| 6.6 Bordereaux validation and exception chasing: Test data quality, identify defects, and push exceptions back to the operational source. | Generated bordereaux, validation rules, and exception reports. | £7 (~6m FinOps, 4m Ops) | £1 Rationale: Automated bordereaux generation removes formatting labour, but exception management stays important until upstream data is clean. | £6 | |
| 6.7 Close pack preparation: Produce the monthly close support pack from the finance outputs. | Ledger outputs, settlement files, and close checklists. | £2 (~3m FinOps) | £1 Rationale: Structured close routines reduce manual pack-building and make the back office more auditable. | £1 | |
| 6.8 IPT support: Prepare the tax support and IPT documentation linked to the accounting close. | Tax logic, policy records, and finance outputs. | £2 (~2m FinOps) | £0 Rationale: Structured tax logic removes routine manual preparation on standard cases. | £2 | |
| 6.9 Control evidence pack: Assemble the audit-ready evidence pack that supports finance and carrier controls. | Control checklists, finance outputs, and reconciliation evidence. | £4 (~1m FinOps, 2m Lead) | £1 Rationale: Automated evidence capture reduces manual collation while leaving humans in charge of sign-off. | £3 | |
| Subtotals (annualised) | Volume: 10,000 bound or changed items /yr | £510,000 | £90,000 | £420,000 | |
| Grand total (annualised) | £5,398,000 | £1,304,000 | £4,094,000 |