The terms SaaS and AI founders actually need to understand.

The accruals principle is foundational to UK GAAP (FRS 102) and US GAAP accounting. It ensures financial statements reflect economic activity in the period it occurred. For a SaaS business with monthly subscriptions billed annually in advance, accruals work means recognising one-twelfth of the contract value as revenue each month, regardless of when the cash was actually collected. The opposite of accruals accounting is cash accounting, which only recognises revenue when received and expenses when paid.

Traditional SaaS unit economics measure customer acquisition cost (CAC) against lifetime value (LTV). For AI businesses, the calculation gets more complex because serving each customer incurs variable cost — typically foundation model API calls, GPU compute, or both. A customer paying £500/mo who consumes £200/mo in compute has very different unit economics from a customer paying the same who consumes £20/mo. Boards and investors increasingly want to see AI unit economics tracked separately from headline ARR figures.

ACV is a foundational SaaS metric used to compare contracts of different lengths on a like-for-like basis. A £30,000 three-year contract has an ACV of £10,000. ACV differs from ARR (Annual Recurring Revenue) in that ACV is calculated per contract, while ARR is calculated across the entire customer base. ACV is particularly useful in pipeline reporting and sales compensation, where contract length needs to be normalised to assess deal value consistently.

For AI businesses, the right classification depends on the business model. If your customer pays you because you serve them an AI response and serving that response costs you an API call, the API spend is directly tied to revenue and sits in COGS. If the API spend is for internal productivity tooling or product development, it’s an operating expense. Specific R&D use cases may qualify for R&D tax relief, though most foundation model usage doesn’t meet the technological uncertainty test. Getting the classification right affects gross margin, investor reporting, and tax position.

For SaaS businesses, ARR is a key operational metric — the value of contracted recurring revenue annualised. But under UK GAAP (FRS 102) and US GAAP (ASC 606), revenue must be recognised as performance obligations are satisfied. A 12-month annual contract billed upfront generates cash on day one but revenue spread evenly across twelve months. Getting this right matters for accurate gross margins, investor reporting, and avoiding restatements at audit. Many SaaS founders confuse ARR (the metric) with recognised revenue (the accounting reality), and the difference becomes material during due diligence.

ASC 606 (Accounting Standards Codification Topic 606) requires revenue to be recognised when performance obligations are satisfied — not when cash is received. For SaaS companies, this means subscription revenue is recognised over the service period, not on the contract signing or billing date. The standard is conceptually similar to IFRS 15, which is the international equivalent that underpins FRS 102’s revenue rules in the UK. SaaS companies operating across US and UK entities need to apply both consistently for group reporting.

For disposals from 6 April 2025 to 5 April 2026, the BADR rate is 14%. From 6 April 2026, the rate rises to 18%. The lifetime limit remains £1 million. For share disposals to qualify, the individual normally needs to hold at least 5% of ordinary share capital, 5% of voting rights, and relevant economic rights for at least two years, and must be an officer or employee of the company. Formerly known as Entrepreneurs’ Relief before the 2020 rename. See HMRC’s BADR helpsheet HS275 for full conditions.

Under UK law, every company must identify and record its beneficial owners (Persons of Significant Control, or PSCs) on the PSC register and at Companies House. A beneficial owner is typically someone who holds more than 25% of shares or voting rights, or otherwise exercises significant influence over the company. The concept is critical for anti-money-laundering compliance and is also relevant for tax purposes — particularly when assessing residency of company controllers for tax-treaty purposes.

Delaware C-Corps are the default structure for US-headquartered startups seeking venture funding. They allow for the issuance of preferred stock with rights and preferences expected by VCs, support employee stock option plans cleanly, and benefit from Delaware’s mature corporate law. For UK founders expanding to the US, setting up a Delaware C-Corp parent (with a UK subsidiary) is one common structuring path — though the right approach depends on tax residency, IP location, and where revenue is generated.

For traditional SaaS businesses, COGS typically includes hosting/infrastructure costs, third-party software licences integrated into the product, payment processing fees, and customer success staff time. For AI businesses, COGS expands to include foundation model API costs (when directly tied to customer-facing inference), GPU compute for serving requests, and other variable AI infrastructure costs. Correct COGS classification is essential for accurate gross margin reporting and is one of the most common areas where AI companies misclassify costs.

Every UK limited company must file specific information with Companies House: annual confirmation statements, statutory accounts, changes to directors or share capital, and details of persons of significant control. The information held at Companies House is publicly accessible, which is meaningful for founders thinking about privacy or competitive considerations. Companies House is distinct from HMRC — Companies House holds company structural information; HMRC handles tax.

When AI companies buy their own GPU hardware (rather than renting cloud compute), the spend is typically capital expenditure — the hardware is an asset that delivers value over multiple years. UK tax law allows capital allowances against this expenditure, and recent rules around full expensing have made GPU purchases particularly tax-efficient. The treatment differs significantly from on-demand cloud GPU usage, which is revenue expenditure deducted in the period incurred. Reserved cloud capacity and long-term commitments sit in a middle ground that depends on the specific contractual arrangement.

A CFC is broadly a non-UK company controlled by UK persons. The rules are designed to tax profits artificially diverted from the UK back to UK controllers. Several exemptions may apply — including the low profits exemption, low profit margin exemption, excluded territories exemption, and tax exemption. For AI and SaaS founders, CFC issues most commonly arise where IP, billing, cash collection, or engineering activity sits in overseas companies controlled from the UK. The inverse direction (UK parent with foreign subsidiary) is where CFC rules typically apply. See HMRC’s International Manual on CFCs for the detailed framework.

Corporation Tax is calculated on taxable profits — accounting profits adjusted for tax-specific add-backs and deductions. The UK main Corporation Tax rate is 25%. The small profits rate is 19%. Marginal relief can apply where profits fall between the lower and upper limits, commonly £50,000 and £250,000, subject to associated company adjustments. Corporation Tax returns (form CT600) are filed annually with HMRC, typically due nine months after the company’s accounting period end. R&D tax relief, capital allowances, and other reliefs are claimed within the Corporation Tax computation. See the Corporation Tax rates legislation for the statutory framework.

The Crossing Roadmap turns a chosen expansion route into a sequenced plan rather than a list of decisions — each step has a date, an owner, and an explicit gate that must be cleared before the next step begins. It is built to be used at board meetings and in investor conversations, where founders need to show not just where the company is going but in what order it will get there. The Roadmap is a living document, updated as circumstances change — a funding round closes, a founder relocates, a customer signs — rather than a one-off plan that goes stale.

US presence becomes an individual reality before it becomes a corporate one, bringing personal residency questions — such as US day-count thresholds — to the front.

CAC is one of the foundational SaaS unit economics metrics. A healthy CAC depends on the lifetime value of the customer (LTV) — the rule of thumb is that LTV should be at least 3x CAC for a sustainable business. CAC payback period (how long it takes for a customer’s contribution margin to repay their CAC) is the related metric many boards prefer because it’s harder to manipulate. For AI businesses, CAC calculations should also account for the variable costs of serving the acquired customer, particularly compute and API spend.

The catalyst sits upstream of any internal strategy: the customer is asking, and the company is at the question rather than the answer.

When a SaaS customer pays upfront for an annual subscription, the cash hits the bank but the revenue cannot be recognised immediately under UK or US GAAP. Instead, the unrecognised portion sits as deferred revenue (also called unearned revenue or contract liability) on the balance sheet. As each month of service is delivered, a portion moves from deferred revenue into recognised revenue. Deferred revenue is one of the most scrutinised items in SaaS due diligence because it represents committed but unearned revenue — a strong indicator of future earnings.

The Defensibility Brief is the output of the Structure stage of the Crossing Method, and it is signed by a named professional rather than left as an anonymous file. It is written in commercial language so a non-specialist reader — a board member, an investor, an acquirer’s counsel — can understand why the structure is the way it is. It is the artefact that makes a structure defensible to HMRC, the IRS, and diligence teams, because it sets out the reasoning at the time the decisions were made rather than reconstructing it years later under pressure.

The Delaware flip is common for US-investor-led rounds, where the lead VC wants a familiar Delaware C-Corp at the top of the structure. It is rarely a tax-neutral move for UK founders — the flip can crystallise exit charges, end EIS relief for existing investors, and create EMI scheme consequences for option holders. The timing of a flip (before or after a round, before or after material revenue, before or after a founder relocates) materially changes its cost, which is why it is treated as a sequenced decision rather than a default.

The treaty determines which country has primary taxing rights over different types of income (dividends, interest, royalties, business profits) and provides mechanisms for relieving double taxation. For SaaS and AI founders operating across both jurisdictions — UK individuals with US subsidiaries, US individuals with UK operations, or anyone with UK-US cross-border revenue — the treaty is the foundational document. Treaty positions need to be considered alongside domestic tax law in both countries, and treaty benefits often require specific elections or filings.

EIS gives qualifying investors 30% Income Tax relief, potential CGT deferral, CGT exemption on qualifying disposals, and loss relief. The company must meet detailed conditions, including qualifying trade rules, gross assets tests, employee headcount limits, age-since-first-commercial-sale restrictions, risk-to-capital tests, and independence requirements. Advance assurance from HMRC is commonly sought before fundraising to confirm the company qualifies. See HMRC’s EIS helpsheet HS341 for the investor relief framework.

EIS continuity is particularly relevant when a US flip or other restructure might trigger an EIS qualifying disposal and lose investors their relief. EIS relief is held by individual investors, not the company — so a structural change made for commercial reasons can claw back relief that investors have already claimed, creating an unexpected tax bill and a difficult conversation. Checking EIS continuity before a restructure, rather than after, is what separates a clean expansion from one that damages the cap table’s most supportive shareholders.

EMI allows qualifying companies to grant tax-advantaged share options to employees. Current core limits include an individual option grant limit and an overall company option limit, plus qualifying company tests around gross assets, employees, and trade. From 6 April 2026, HMRC guidance reflects increases to certain EMI company limits — relevant for scaling SaaS and AI companies that may have outgrown the previous thresholds. Employees must normally meet the working time requirement of at least 25 hours per week, or — if less — 75% of their working time spent on the company’s business. EMI notifications must be made to HMRC within the required deadline to preserve the tax-advantaged treatment. See HMRC’s EMI qualifying company guidance and the working time requirement amendment.

The Exposure Score is the output of the Risk stage of the Crossing Method, and it answers a different question from “what should we do next” — it answers “what have we already done.” Each of the six dimensions is read as green, amber, or red, so a founder can see at a glance where exposure has already begun. Founders often discover that exposure started long before they thought it had — a contractor closing US deals, a founder’s travel pattern, IP that drifted across a border — which is precisely why the Score is run before any routes are compared.

Flip-ups are common for UK SaaS and AI startups raising from US-based VCs who prefer (or require) a Delaware C-Corp investment vehicle. The transaction typically involves UK shareholders exchanging their UK shares for shares in a new US holding company, with the UK entity becoming a subsidiary. The transaction has significant tax, securities, and operational implications — particularly around CGT exposure for UK founders, US tax residency rules, transfer pricing for ongoing operations, and the practical complexity of US compliance. The right timing for a flip-up (before vs after a funding round, before vs after material revenue, before vs after a founder relocation) is one of the most consequential structural decisions a UK SaaS founder makes.

For most AI companies, foundation model API spend is the largest single AI-specific cost line. Pricing typically follows a per-token model (input tokens + output tokens, with different rates for each), with reserved capacity and committed-use discounts available at higher volumes. The accounting treatment depends on what the spend is for — customer-facing inference is typically COGS, internal productivity tooling is OpEx, experimental development can sometimes qualify as R&D. VAT treatment also needs careful consideration since most foundation model providers are based outside the UK, triggering reverse charge rules.

Founder residency is not a matter of where someone feels based — it is a technical test, and two tests apply at once: the UK Statutory Residence Test and the US substantial presence test. Each country runs its own calculation on its own facts, so it is entirely possible to be resident in both at the same time. Getting it wrong creates dual-residency, an unexpected personal tax bill, and treaty tie-breaker arguments that are expensive to run after the fact rather than planned for in advance.

FRS 102 applies to UK private companies that aren’t required to use full IFRS. It’s a simplified standard derived from IFRS, with specific UK adaptations. For SaaS and AI businesses, the most relevant FRS 102 sections cover revenue recognition (Section 23), intangible assets and development costs (Section 18), lease accounting (Section 20), and consolidation rules for groups. FRS 102 is updated periodically — most recently with material changes to revenue and lease accounting that affect SaaS businesses specifically.

GPUs and server hardware used in the trade will normally qualify as plant and machinery. New qualifying main-rate plant may qualify for full expensing, which gives a 100% first-year allowance. The Annual Investment Allowance (AIA) is currently £1 million. Integral features in data centre infrastructure — such as certain electrical systems, cooling, or power infrastructure — may fall into the special rate pool with a 50% first-year allowance for new qualifying expenditure. Second-hand equipment does not qualify for full expensing but may qualify for AIA or writing-down allowances. Disposals of GPUs can trigger balancing charges, particularly relevant where there’s an active resale market and full expensing has been claimed. See Full expensing, the Annual Investment Allowance guidance, and the Capital allowances manual for detail.

Gross margin is one of the most-watched SaaS and AI business metrics because it indicates the underlying unit economics of the business. Traditional SaaS businesses typically achieve 70–90% gross margins because hosting costs are relatively low. AI businesses, by contrast, often see compressed gross margins (30–60% in some cases) because inference costs scale with usage. Investors increasingly want to see AI-specific gross margin reported separately, with foundation model API spend and GPU compute costs broken out from traditional infrastructure costs.

The structural question is how to employ someone in the US, rather than whether a US entity is needed for other reasons.

HMRC investigations span a spectrum from routine compliance checks (most common) to specialist investigations of suspected fraud (rare). For SaaS and AI companies, the most common triggers for HMRC scrutiny are R&D tax credit claims (which have seen significantly increased scrutiny in recent years), VAT positions on cross-border digital services, and large or unusual deductions. Most investigations are resolved through documentation and explanation rather than adversarial proceedings, but the time and cost can still be material. HMRC investigation insurance covers the professional costs of representation during an enquiry.

Holding company structures are common in international expansion, group reorganisations, and tax planning. For SaaS and AI founders, a UK HoldCo with a US C-Corp subsidiary (or vice versa) is one common structuring path. The right structure depends on where IP sits, where revenue is generated, where founders reside, and the long-term strategic plan. Holding company structures also enable separating operating risk (held in the operating subsidiary) from accumulated value (held at HoldCo level), which has implications for asset protection and tax efficiency on exit.

IFRS 16 replaced the previous operating lease vs finance lease distinction with a single lessee accounting model. The UK equivalent under FRS 102 has more flexibility but is broadly aligned. For AI companies, IFRS 16 considerations matter for office leases, GPU equipment leases, and — increasingly — long-term cloud compute reservations that may meet the definition of a lease. Whether a multi-year cloud GPU reservation is a lease (capitalised) or a service contract (expensed as incurred) is a real and consequential accounting judgement.

For England, Wales, and Northern Ireland, the standard Personal Allowance is £12,570. The main Income Tax bands are 20% (basic rate), 40% (higher rate), and 45% (additional rate) — applied to income above the personal allowance. The dividend allowance is £500 for 2025/26, with dividend tax rates set separately from non-dividend income. Scotland operates separate income tax rates and bands for non-savings, non-dividend income. Savings allowances apply separately. Share option exercises can create employment income and NIC charges unless a tax-advantaged regime such as EMI applies. See Income Tax rates and allowances for current figures.

The company is being structurally pulled by capital rather than by customers or hiring.

An individual passes the Substantial Presence Test (and is therefore treated as a US tax resident) if they’re present in the US for at least 31 days in the current year AND 183 days across a weighted three-year calculation (full days this year + 1/3 of days last year + 1/6 of days the year before). For UK founders spending significant time in the US for business, this calculation matters enormously — passing the test exposes worldwide income to US tax, potentially without treaty relief depending on the situation. Tracking days carefully and structuring trips is a meaningful operational consideration for cross-border founders.

For non-US founders who own shares in a US C-Corp, receive US-source income, or need to file a US tax return for any other reason, an ITIN is typically required. The application process (Form W-7) requires either certified copies of identification documents or in-person verification through an IRS-authorised acceptance agent. ITINs expire if not used on a federal tax return for three consecutive years, requiring renewal. Getting the ITIN process right is one of the practical first steps when UK founders flip-up to a US C-Corp structure.

LTV is the partner metric to Customer Acquisition Cost (CAC). The most common formulation is Average Revenue Per Customer × Gross Margin / Churn Rate, giving the expected contribution margin a customer delivers over their expected tenure. Some businesses calculate gross LTV (revenue without margin) which can mask underlying economics. For AI businesses, LTV calculations should account for variable serving costs as the customer scales — a customer doubling their AI usage doesn’t double their LTV if compute costs scale proportionally.

LLCs differ from C-Corps in their default tax treatment — LLCs are typically pass-through entities (members report income on their personal returns) while C-Corps are separately taxed entities. For venture-backed SaaS and AI startups, LLCs are usually unsuitable because most institutional investors require C-Corp structure. However, LLCs are common for solo founders, consultancies, and businesses not seeking venture funding. For UK founders, owning a US LLC has specific UK tax implications (HMRC’s treatment of US LLCs is complex and case-specific) that need careful consideration before formation.

Where statutory accounts are an annual external compliance requirement, management accounts are an internal operational tool — typically monthly or quarterly, focused on operational decision-making rather than statutory compliance. For SaaS and AI businesses, management accounts usually include P&L by department, balance sheet, cash flow forecast, ARR/MRR tracking, customer metrics, and budget-vs-actual analysis. The format and depth varies based on business stage and investor reporting requirements. Investors typically expect monthly management accounts within 15–30 days of month-end.

For accounting periods beginning on or after 1 April 2024, the merged R&D expenditure credit scheme replaced the previous split between the SME R&D scheme and the large company RDEC scheme for most companies. The headline credit is 20% of qualifying R&D expenditure. Because the credit is taxable, the net cash value is normally 15% where the company pays Corporation Tax at 25%, or 16.2% at the 19% small profits rate. A separate Enhanced R&D Intensive Support route (ERIS) remains for qualifying loss-making R&D-intensive SMEs where qualifying R&D expenditure is at least 30% of total expenditure — the enhanced payable credit rate is 14.5%. See HMRC’s guidance on the merged scheme and ERIS and the ERIS intensity threshold detail.

For AI companies, model training is one of the most material cost categories and one of the most complex from an accounting perspective. The cost can include GPU compute (owned or rented), data acquisition or annotation, ML engineer salaries (apportioned to qualifying R&D activity), foundation model API costs (for fine-tuning), and third-party services. The right accounting treatment depends on whether the activity qualifies as R&D (potentially eligible for tax relief), whether costs should be capitalised as development costs under FRS 102 Section 18, and whether the resulting model is treated as an internally generated intangible asset. Most AI companies underclaim R&D relief on qualifying training work because the apportionment is complex.

The defining feature is not which catalyst leads, but that the company is being pulled from several directions simultaneously.

Employer and employee NIC rates and thresholds are set annually. Employer NIC is particularly relevant for payroll cost, benefits in kind (Class 1A NICs), PAYE settlement agreements (Class 1B), and certain share option exercise events. Self-employed individuals pay Class 2 and Class 4 NICs through Self Assessment. Dividends are not subject to NIC — one of the structural reasons founders often take income as a mix of salary and dividends rather than salary alone. Employment Allowance may reduce employer NIC for eligible employers. See HMRC’s rates and thresholds for employers 2025-26 for current figures.

NRR is calculated as: (Starting MRR + Expansions − Contractions − Churn) / Starting MRR, expressed as a percentage. NRR above 100% means existing customers are growing in value despite some churn — a strong signal of product-market fit and customer expansion. Best-in-class SaaS businesses achieve NRR of 120%+. NRR is increasingly cited as the most important SaaS metric because it captures the compounding effect of expansion revenue, which dominates long-term enterprise value. For usage-based AI businesses, NRR can be volatile because customer usage scales with their own growth.

Every UK employer must operate PAYE, calculating tax and NI on each pay run and remitting to HMRC monthly. PAYE submissions happen in real time through Real Time Information (RTI) — each pay run is reported to HMRC at or before the time of payment. For SaaS and AI startups, PAYE complexity grows with: equity participation (share option exercises trigger PAYE in some cases), benefits in kind (private health, gym memberships, employer-provided equipment), and international employees (UK PAYE for non-UK residents, treatment of overseas workdays). Errors in PAYE filings are one of the most common causes of HMRC contact for early-stage companies.

Under both ASC 606 (US) and FRS 102 (UK), revenue recognition is tied to satisfying performance obligations rather than receiving cash. For SaaS contracts, the typical performance obligation is providing ongoing access to the software platform over the subscription period — revenue is recognised evenly across that period. But complex contracts can have multiple performance obligations: software access (recognised over time), implementation services (recognised on completion), training (recognised when delivered), each with separate timing. Identifying and allocating value to performance obligations is one of the most consequential accounting decisions for SaaS businesses.

PE is the threshold concept for whether a foreign company’s activities in a country are sufficient to create taxing rights for that country. Traditional PE definitions include offices, branches, and dependent agents who habitually conclude contracts. Modern interpretations under treaties also consider digital activities, though the rules are still evolving. For SaaS and AI companies, the most common PE risks are: sales personnel based in another country who close deals there, technical staff working from another country, and contractors who effectively operate as part of the company. PE creation can result in unexpected tax filings, transfer pricing requirements, and corporate tax exposure in the other jurisdiction.

The live questions are what follows the move: transfer pricing, IP migration, intercompany agreements, and defensibility.

The work at this stage is framing the questions clearly before any single pressure starts driving decisions.

When a company pays for an annual insurance policy upfront, the cash leaves the business immediately but the cost relates to twelve months of coverage. Prepayments accounting holds the unused portion as a current asset, releasing one-twelfth into the P&L each month. The same principle applies to prepaid software licences, rent paid in advance, prepaid professional fees, and reserved capacity payments. For AI companies, prepaid cloud reservations and committed API spend often sit as prepayments — the treatment can be material for understanding actual operating costs vs cash outflows.

A Person of Significant Control is anyone who holds more than 25% of shares, more than 25% of voting rights, the right to appoint or remove the majority of directors, or otherwise exercises significant influence or control over the company. PSC information is publicly accessible on the Companies House register. For startup founders, PSC reporting becomes more complex as investor classes are introduced (preferred shares with super-voting rights), trust structures hold shares, or corporate shareholders are themselves controlled by individuals further up the chain. Maintaining the PSC register accurately is a statutory compliance requirement separate from filing annual confirmation statements.

R&D tax relief applies where a company seeks an advance in science or technology by resolving scientific or technological uncertainty that could not readily be resolved by a competent professional in the field. Claims require qualifying projects, qualifying expenditure, correct treatment of contracted-out work, and submission of an Additional Information Form. Under the merged R&D expenditure credit scheme (for accounting periods from 1 April 2024), most companies claim a 20% above-the-line credit, with ERIS available for qualifying R&D-intensive loss-making SMEs at a 14.5% enhanced payable credit. AI claims need particular care around prompt engineering, foundation model API costs, cloud compute apportionment, and competent professional evidence. See HMRC’s R&D tax relief collection, the R&D software guidance, and the Additional Information Form requirements.

Major cloud providers offer reserved instances, savings plans, and committed use discounts that significantly reduce the per-hour cost of compute in exchange for upfront commitments (often 1–3 years). For AI companies running material training or inference workloads, reserved capacity can reduce costs by 30–60% versus on-demand pricing. The accounting treatment is less straightforward than on-demand spend — depending on the structure, reserved capacity might be a prepayment, a service contract, or in some cases a lease under FRS 102 or IFRS 16. Multi-year dedicated GPU contracts with providers like CoreWeave often raise lease accounting questions specifically.

The Routes Matrix is the output of the Options stage of the Crossing Method, and its job is to replace a default assumption — usually “flip to Delaware” — with a scored decision. Each of the five routes is assessed against the company’s actual facts rather than a generic template, so the ranking reflects this company’s tax position, cap table, and commercial timeline. The result is not a recommendation in the abstract but a comparison a founder and board can interrogate route by route.

Under FRS 102 (UK) and ASC 606 (US), SaaS revenue must be recognised as the customer receives the benefit of the software service, not when cash is collected. For a standard annual SaaS contract billed upfront, this means one-twelfth of the contract value is recognised each month over the year. Multi-element contracts (subscription + implementation + training + support) require revenue to be allocated across each performance obligation and recognised separately. The complexity increases with usage-based pricing, hybrid pricing models, milestone-based deliverables, and contracts with significant up-front discounts.

SEIS gives qualifying early-stage investors 50% Income Tax relief, CGT exemption on qualifying disposals, loss relief, and CGT reinvestment relief, subject to detailed investor and company conditions. The scheme is designed for the earliest-stage companies — with strict limits on gross assets at the point of investment, age since trading commenced, employee numbers, and qualifying trade restrictions. SEIS is commonly used before EIS in a startup’s funding journey, with companies typically moving from SEIS rounds (first £250,000 or so) to EIS rounds as they scale. See HMRC’s SEIS helpsheet HS393 for current rules.

Most UK employees have their tax handled entirely through PAYE and never file a tax return. Self Assessment is required for: self-employed individuals, company directors (typically), individuals with significant investment income or capital gains, individuals with rental income above thresholds, and high earners with income over £100,000. The tax year runs 6 April to 5 April, with returns due online by the following 31 January. For SaaS and AI founders, Self Assessment captures dividend income, share option exercises, capital gains on share disposals, and other items not handled through company payroll.

Standard founder and employee vesting in venture-backed startups is four years with a one-year cliff: no shares vest in the first year, then 25% vest after the one-year anniversary, with the remaining 75% vesting monthly or quarterly over the following three years. Vesting protects against early departure by ensuring contributors earn their equity over time. For founders, vesting is typically negotiated with investors as part of a funding round — pre-investment, founders may hold fully vested shares; post-investment, those shares usually get put on a vesting schedule (with credit for time served). Vesting interacts with EMI schemes, tax treatment of exercise events, and exit scenarios in ways that require careful structuring.

Statutory accounts (also called annual accounts or financial statements) are prepared under FRS 102, FRS 105, or full IFRS depending on company size and circumstances. They typically include a balance sheet, profit and loss account, cash flow statement, and notes. Filing requirements vary by company size — small companies file abbreviated accounts at Companies House (publicly accessible) but full accounts with HMRC. Statutory accounts are distinct from management accounts (used internally) and are the official record of financial performance that creditors, suppliers, and the public can rely on.

Where Companies House holds summary information, the statutory register is the underlying detailed record that the company itself maintains. The register must be kept up to date and is required to be accessible for inspection. Components include: register of members (shareholders), register of directors, register of secretaries, register of persons of significant control, register of charges, register of debenture holders (if any), and minutes of meetings. Maintenance of the statutory register is often overlooked by founders but is a real compliance requirement — failure can lead to fines and complications during due diligence.

The SRT operates through a series of tests applied in order: automatic UK tests (which definitively make someone UK resident), automatic overseas tests (which definitively make someone non-resident), and the sufficient ties test (which considers a combination of days in the UK and connections like family, accommodation, work, and prior residence). For founders considering relocation — whether to the US for tax efficiency, to lower-tax jurisdictions, or returning to the UK after time abroad — the SRT is the foundational document. The calculation is highly fact-specific and requires careful day-counting and documentation.

Many AI products charge customers per token consumed (per million input tokens, per million output tokens), per API call, or per other usage metric. Under ASC 606 and FRS 102, this revenue is typically recognised as the usage occurs — the performance obligation is satisfied at the moment of service delivery. This differs from subscription-based SaaS revenue, which is recognised evenly over the contract period. Tokenisation pricing creates volatile monthly revenue patterns and makes ARR-style metrics less meaningful, often requiring custom unit economics frameworks. Some AI companies use hybrid pricing (base subscription + usage overages) which requires careful allocation of revenue between performance obligations.

When a UK company sells services to its US sister company (or vice versa), the price charged must be the price an unrelated party would charge — the arm’s length principle. Transfer pricing rules prevent companies from shifting profits to lower-tax jurisdictions by manipulating intra-group pricing. For SaaS and AI groups with UK and US entities, transfer pricing affects: IP licensing fees between entities, shared service charges (engineering, R&D, management), cost-plus arrangements, and revenue allocation when customers are served from one entity but contract with another. Documentation requirements are significant, with detailed studies often required to support intra-group pricing positions.

Trigger Classification is the output of the Catalyst stage of the Crossing Method, and it does something deceptively simple: it names the real trigger. Naming the trigger separates genuine commercial drivers — a signed customer that needs a US contracting entity — from borrowed urgency, where a founder is reacting to what a peer company did rather than to their own facts. Once the trigger is named, the typical structural response that follows it can be assessed honestly rather than assumed.

A US permanent establishment triggers US federal and state filing obligations — corporate income tax returns, and potentially state-level filings wherever nexus is created. A UK SaaS company can create one without realising, for example through a US-based salesperson who habitually closes deals, a US contractor acting as a dependent agent, or a fixed place of business that grows past an incidental presence. Because the threshold is activity-based rather than headcount-based, the exposure often exists well before a founder considers the company to have a US operation.

Usage-based pricing has become particularly common for AI and infrastructure SaaS businesses where the cost of serving customers varies significantly with usage. Examples: per-token AI APIs (OpenAI, Anthropic), per-API-call usage at Twilio or Stripe, per-compute-hour cloud services. The accounting and operational implications are significant: revenue is recognised as usage occurs (more volatile than subscription revenue), ARR calculations become less reliable predictors, customer acquisition and retention metrics need adjustment for usage variability, and unit economics need to be tracked at the per-transaction or per-customer level rather than per-seat.

The UK standard VAT rate is 20%, the reduced rate is 5%, and the zero rate is 0%. The VAT registration threshold is £90,000 of taxable turnover in any rolling 12-month period; the deregistration threshold is £88,000. VAT returns are normally submitted quarterly, though monthly and annual schemes exist. All VAT-registered businesses must comply with Making Tax Digital (MTD) for VAT — digital record-keeping and direct API submission to HMRC. Digital services to consumers and businesses, and cross-border services more generally, require careful place-of-supply analysis and may trigger reverse charge obligations. See VAT rates, the VAT registration threshold guidance, MTD for VAT, VAT on services from abroad, and VAT Notice 741A on place of supply for detailed rules.

When a UK business buys services from a non-UK supplier (foundation model APIs from OpenAI, Anthropic, cloud services from AWS US, etc.), the supplier doesn’t typically charge UK VAT. Instead, the UK business must account for the VAT itself through the reverse charge — adding output VAT and (where recoverable) input VAT in the same return. The net cash effect is usually neutral, but failure to apply the reverse charge correctly is a common compliance error for SaaS and AI companies with significant overseas service spend. The reverse charge applies to both B2B services and certain digital services received from outside the UK.