Hi operators, special mid-week drop! We've put together a term sheet to simplify and standardise debt fundraising terms for AI services companies: the SAIL (Standard Agreement for Invoice Lending). If you want us to go deeper on specific clauses or the rationale behind them, drop a comment below!
Friday issue plus our pricing survey lands as normal - watch for it!
-Daria
Demystifying debt for AI Services operators
A few months back we covered financing strategies for AI services businesses: traditional funding paths, different types of providers, and what founders in AI enablement should keep in mind when engaging with each. Part 1 covers dilutive/equity funding and part 2 covers non-dilutive/debt funding.
The debt instalment got a strong response. Not hard to see why. AI services companies bill customers from day one. There's no multi-year R&D phase to fund before revenue arrives, and with that no absolute need for capital. Many founders we speak with don't have the appetite to raise equity, even when the business could genuinely use capital to grow faster. Debt is a different conversation: you access capital, accelerate, and don't put a valuation on the business and don’t lose full control of the business.
The question is which kind of debt. Term loans require hard assets or a trading history most young AI services firms don't have. Revenue-based financing assumes predictable recurring revenue, which doesn't fit consulting or implementation businesses with lumpy, project-based income. Venture debt typically requires institutional equity investors as a backstop. Invoice discounting is different. The collateral is already sitting on your balance sheet: the invoices you've raised against work you've delivered. The diligence focuses on your customers, not your business as a whole.
The catch is that traditional lenders are still not set up to underwrite AI services businesses properly. Frameworks built for SaaS or traditional SMEs don't translate. And among the lenders who do engage, terms vary enormously.
The terms below are our best attempt to set a standard for what a debt agreement between an AI services company and a lender should look like. They're the result of conversations with founders, accountants, lawyers, and traditional lenders, as well as several deals we've done ourselves.
Introducing SAIL: Standard Agreement for Invoice Lending
All subscribers can download the term sheet here. Key terms are explained below. If you'd like us to write a full guide, leave a comment!
Fees
Invoice discounting facilities can include up to six different charges. The main ones to know:
Annualised discount rate: interest on what you've actually drawn down. Typically 8–12% for established businesses; higher for AI services companies given shorter track records and more complex delivery patterns.
Facility fee: a standing charge for having the line available, whether you use it or not.
Legal fees: covers external legal work (facility agreement, security documents, contract review). Passed through to the borrower, typically capped around £5,000. Sometimes bundled together with the arrangement fee, but fundamentally covers two different things.
Arrangement fee: the lender's charge for their own internal effort: diligence, credit analysis, structuring. Traditional lenders charge 1–2% of the facility limit upfront. We waive this to keep the economics founder-friendly.
Per-invoice admin fee: common in US and fintech models, less standard in the UK. Covers the effort going into assessing and verifying each invoice that goes into the facility.
Override premium: a pricing uplift when a lender makes an exception to standard eligibility criteria. Gives flexibility without changing the base rate.
Fees are the main thing to negotiate in any facility agreement — and watch for lenders bundling arrangement fees with legal fees, since they're covering two different things.
Eligible Invoices
The core rule is simple: you don't lend against work that hasn't been delivered. A signed contract doesn't guarantee payment if the service hasn't happened yet — clients can cancel, dispute scope, or change their mind. So eligibility always starts with: has the work been done?
Fully eligible — service fully delivered, signed contract from both parties, debtor passes credit check. Advance rate: 85% of invoice value.
Eligible – discounted — partially delivered, with the client confirming the delivery percentage directly to the lender. Advance rate: proportional to confirmed delivery (e.g. 60% delivered = 60% × 85% = 51% of invoice value). Or fully delivered but with a purchase order rather than a signed contract.
Watchlist — delivered, but documentation gaps (e.g. unsigned contract). No advance until resolved, but can be reclassified quickly once the paperwork is in order.
Ineligible — not yet delivered, future-dated, disputed, chronic non-payer, or high cross-border enforcement risk.
This is where working with a lender who understands AI services is a significant advantage. A high-street bank won't lend against a services receivables book because they have no framework for assessing whether the work has actually been done. They can't evaluate a milestone sign-off on an AI implementation project or judge whether 60% of a training programme counts as delivered. Traditional lenders are set up for physical goods or simple recurring contracts. Partial delivery assessments and project milestone structures aren't something they're equipped to underwrite.
A few other things specific to AI services businesses:
Payment terms: large enterprise clients often have 90-day standard terms. If a lender excludes anything 90 days past due date, you may effectively be waiting 180 days from invoice date. A specialist lender builds this into how they assess overdue status.
Concentration: if one client represents a large share of your receivables, most lenders reduce the advance rate on that exposure. Aim for a diversified debtor pool; 20–30% per debtor is standard market guidance.
Minimum invoice size: small invoices (under ~£10–15k) often aren't worth the processing overhead for either side and may be excluded from the eligible pool.
Credit checks on debtors: blue-chip enterprises are assessed on track record; smaller or less well-known clients will need a formal credit check. This is standard and low-cost, but it needs to be factored in.