Seventy-eight percent of wholesale distribution businesses are planning to adopt or expand artificial intelligence in their enterprise resource planning systems, embedding predictive algorithms into demand planning, inventory management, and procurement cycles. This adoption pattern reflects a broader trend: mid-market businesses across manufacturing, services, retail, and construction are rapidly advancing their operational sophistication.
The paradox? While these businesses become more algorithmically sophisticated in their operations, their financial service providers struggle to keep pace with algorithms of their own.
Businesses generating $10 to $50 million in annual revenue occupy a structural gap in financial services. Retail banking’s algorithmic underwriting can’t accommodate operational complexity – covenant structures tied to rental roll performance, staged construction drawdowns, working-capital lines calibrated to seasonal cash cycles. Yet transaction sizes, typically $10–30 million in facility value, can’t justify corporate banking’s relationship manager economics.
Three distinct models have emerged – technology platforms scaling upward, institutional professionals in boutique practices, and purpose-built banks – and their coexistence suggests fragmentation’s permanent. It’s driven by heterogeneity within the segment itself.
Table of Contents
The Structural Gap
This paradox between operational sophistication and financial service limitations isn’t coincidental – the missing middle exists as a structural consequence of irreconcilable economic constraints. Algorithmic systems require standardisation that mid-market complexity prevents, while relationship banking requires scale that mid-market transactions can’t provide. Retail banking algorithms function by standardising inputs – credit scores, debt-service coverage ratios, asset-based lending formulas. Efficiency requires predictability.
However, mid-market businesses generating $10–50 million require covenant structures responding to business-specific operational dynamics: rental roll fluctuations for property agencies, presales commitments for construction projects, inventory-to-receivables ratios for seasonal retailers. Here’s the irony: these businesses are embedding AI into their operations while their lenders remain stuck with algorithms designed for much simpler cases.
Corporate banking relationship models assume dedicated relationship managers whose annual compensation must be justified by fee revenue per client. A corporate client generating substantial annual fees can support dedicated coverage. A mid-market business generating significantly lower facility fees can’t.
This gap’s structural, not transitional. Thirty-five percent of mid-market businesses have already implemented AI into their processes. As businesses in this revenue band adopt more sophisticated operational systems, their financial complexity increases while transaction scale remains insufficient for traditional corporate banking investment.
This economic tension has created space for specialised providers, each attempting to resolve the sophistication-scale paradox through fundamentally different approaches that can’t converge.
The Algorithmic Efficiency Model
Technology-enabled lenders demonstrate that algorithmic underwriting can achieve deployment scale and attract institutional capital, but their economic model depends on standardisation that defines a ceiling where mid-market customisation begins. Technology-enabled lending’s the first model attempting to serve the missing middle by scaling algorithmic efficiency upward from small business lending. This approach depends on limiting customisation – automated decisioning requires inputs that can be systematically evaluated without manual intervention.
Greg Moshal co-founded Prospa in 2012, building a technology platform that’s deployed over $450 million to small businesses through standardised digital underwriting. The company raised more than $200 million from institutional investors including Carlyle Group, Ironbridge Capital, and AirTree Ventures.
Prospa’s success validates the economic model but requires covenant structures to follow templates, security packages to fit standard categories, and monitoring to occur through automated reporting rather than relationship manager oversight. Try explaining rental roll fluctuations to a lending algorithm sometime – it won’t end well.
Prospa’s success validates that algorithmic efficiency scales profitably for businesses with standardisable borrowing needs but defines the boundary where technology-enabled lending confronts mid-market nuance. Look, ‘standardisable’ simply means your borrowing requirements fit existing templates without manual tweaks – no special covenants, no unique security structures. When businesses require covenant packages tied to rental roll performance or staged drawdowns linked to construction milestones, standardisation confronts limitations. Customisation fundamentally conflicts with the economics of standardisation.
Operational Complexity and Cash-Flow Dynamics
The standardisation constraint becomes particularly problematic when mid-market businesses deploy sophisticated operational systems that generate dynamic, business-specific cash-flow patterns requiring lenders to continuously model how operational variables affect credit risk – analytical work that can’t be systematised into algorithmic thresholds. Businesses implementing predictive demand planning and automated procurement systems generate cash-flow patterns responding to algorithmic business operations. Inventory levels adjust to forecasted sales cycles, receivables timing shifts with automated fulfilment, working capital requirements fluctuate based on system-driven ordering.
This operational complexity creates financial service requirements simple credit algorithms can’t address. Lenders must model how the business’s own algorithms affect cash generation – not just evaluate historical financial statements or apply static credit ratios.
Specific analytical requirements emerge: covenant structures responding to operational cycles rather than calendar periods, monitoring frameworks tracking performance metrics specific to business models, security packages accommodating operational assets that algorithms manage.
When operational sophistication produces cash-flow dynamics requiring ongoing professional judgement about credit risk, businesses move beyond what algorithmic platforms can serve – necessitating institutional banking methodologies that boutique providers deliver.
Institutional Methodology at Mid-Market Scale
Boutique financial services providers applying institutional banking methodologies to mid-market transactions demonstrate the analytical depth this segment requires, but individual analyst capacity creates scalability constraints that prevent this model from achieving the volume algorithmic platforms reach. Institutional banking methodologies involve financial model building with stress testing under various scenarios, covenant calibration responding to business-specific performance drivers, security structures accommodating portfolio dynamics, ongoing monitoring requiring professional judgement rather than automated alerts.
Martin Iglesias, a Sydney-based Credit Analyst at Highfield Private, structured over $30 million in lending facilities for a real estate agency’s portfolio growth. He assembled term debt and working-capital lines against rental roll and property security, setting reporting undertakings around rent roll performance, loan-to-value ratios, and interest cover ratios.
Why does this level of detail matter? Because it’s exactly what separates mid-market lending from clicking ‘approve’ on a small business loan application.
The real estate agency transaction required ongoing monitoring of rental roll fluctuations to assess interest coverage under stress scenarios, covenant calibration responding to lease renewal cycles rather than static balance sheet ratios, security structures accommodating portfolio expansion while maintaining loan-to-value discipline. These aren’t inputs that can be standardised into algorithmic thresholds – they require continuous professional judgement.
The rental roll monitoring, covenant calibration, and security structuring in the real estate agency transaction demonstrate mid-market lending requires ongoing analysis of dynamic performance metrics – illustrating why algorithmic systems designed for static credit scoring can’t replicate the analytical depth distinguishing mid-market from small business lending.
Systematising Relationship Banking
Purpose-built banks represent a third approach, attempting to industrialise relationship banking sophistication through proprietary technology infrastructure. This approach bets covenant analysis and cash-flow assessment can be systematised without losing customisation capacity but requires capital investment that creates its own profitability threshold. Embedding institutional banking sophistication into scalable technology systems rather than relying on individual banker expertise attempts to resolve the tension between analytical depth and operational efficiency through purpose-built infrastructure.
Chris Bayliss, CEO and Managing Director of Judo Bank, has overseen development of a proprietary core banking platform in collaboration with Thought Machine. The bank’s achieved an A$10 billion loan book through a model attempting to embed covenant analysis and cash-flow assessment into systematic processes while preserving relationship-based decisioning.
Judo’s approach represents a bet that relationship banking sophistication – covenant design, security structuring, financial modelling – can be industrialised through purpose-built technology rather than adapted from retail banking platforms. The bank recruits relationship bankers and opens regional locations, pursuing scale through systematised analytical processes without fully automating credit decisions.
Sure, but there’s something contradictory about ‘systematising’ relationship banking – you’re either building authentic relationships or you’re running a system.
Judo Bank’s development of proprietary infrastructure reflects an attempt to resolve the sophistication-scale tension by systematising institutional banking methodologies. However, the loan book must ultimately demonstrate whether industrialised relationship banking achieves profitability justifying the capital investment required to build parallel infrastructure outside major bank systems.
Heterogeneity as Permanent Structure
The persistence of three distinct models – algorithmic scaling, boutique expertise, and industrialised relationship banking – indicates something significant: mid-market businesses don’t constitute a homogeneous segment awaiting optimal solution but rather encompass distinct complexity tiers requiring incompatible economic approaches that won’t converge. The $10–50 million revenue band encompasses algorithmically serviceable businesses with predictable cash cycles and standardised security packages; expertise-dependent businesses requiring bespoke covenant calibration and ongoing performance monitoring; and infrastructure-justifying businesses with sufficient transaction volume or sector concentration to support dedicated banking platforms. Market segmentation meets messy reality.
Prospa’s deployment volume indicates that the algorithmically serviceable sub-segment’s substantial – businesses whose borrowing needs map to template covenant structures and whose operations generate predictable enough cash flows for standardised monitoring.
The real estate agency transaction requiring rental roll analysis and portfolio-specific loan-to-value management exemplifies businesses in the expertise-dependent tier – those whose operational complexity demands ongoing professional judgement rather than automated assessment.
Businesses deploying predictive inventory and demand-planning systems generate cash-flow patterns driven by their own algorithmic operations. As these firms expand algorithmic operations, they push upward in complexity tier regardless of transaction size.
This widens the analytical gap between what retail banking algorithms can assess and what these businesses require.
The Economics of Incompatibility
Economic constraints prevent each model from profitably expanding across complexity tiers, creating permanent boundaries that sustain fragmentation rather than temporary inefficiencies that competition will eliminate. Algorithmic platforms can’t profitably add manual covenant customisation: unit economics depend on standardisation and automated processing. Adding relationship manager oversight destroys the cost structure that enables high-volume deployment.
Boutique practices face capacity constraints: individual analyst expertise can’t scale beyond portfolio limits without diluting analytical depth. Hiring more analysts increases costs without proportionally increasing capacity if transactions require senior-level judgement.
Purpose-built banks face capital barriers: proprietary technology development requires investment that must be justified by transaction volume and profitability thresholds.
Expanding across all complexity tiers would require serving businesses whose economics can’t support the infrastructure cost.
A Fragmented Future
The gap between algorithms and boardrooms isn’t a transitional inefficiency awaiting resolution but a permanent feature sustained by heterogeneity within the mid-market segment itself. Businesses generating $10–50 million require sophistication varying from standardisable to expertise-dependent to infrastructure-justifying, and no single economic model can profitably span this spectrum.
As wholesale distributors embed artificial intelligence into operational systems, the analytical distance between their own algorithmic sophistication and available lending algorithms grows wider, not narrower. The financial services industry keeps searching for that one perfect solution to serve everyone – classic wishful thinking about what’s actually three different markets.
The future of mid-market financial services won’t feature a dominant model emerging from competitive consolidation – fragmentation itself is the equilibrium. Businesses caught between algorithms and boardrooms will remain there not because financial services providers lack innovation but because the missing middle isn’t a cohesive market segment. It’s a collection of sub-segments requiring irreconcilable economic models, each sustainable in isolation but unable to profitably colonise the others’ territory.
As businesses continue embedding AI into their operations, the sophistication gap only widens, reinforcing that this structural feature won’t resolve through competition or innovation. The gap persists because it’s not actually a gap.
It’s three separate markets wearing a trench coat.
