TL;DR

Shawbrook posted £195.5m in underlying pre-tax profit for the first half, against £168.6m a year earlier, and credited part of the 16% rise to AI deployed across its lending operations. Its cost-income ratio fell to 36.4% from 40%.

That ratio is the number worth watching. Plenty of UK firms describe AI programmes; comparatively few point at a specific efficiency measure and attribute movement in it. Shawbrook’s real estate division offers a concrete example, where AI-assisted underwriting is targeting up to a 50% cut in preparation time.

Chief executive Marcelino Castrillo was careful about the attribution, describing the gains as the product of deliberate choices across organisational design, property footprint and technology stack, increasingly helped by AI, with benefits compounding over time. That framing places AI as one contributor to an efficiency programme rather than its cause — a more defensible claim than the sector usually makes, and harder to disprove either way.

The rest of the half-year results give the context. Lending reached £20.1bn, representing 10% annualised growth once originate-to-distribute business is counted. A pair of securitisations worth £1.3bn between them delivered £25.8m in gain on sale. Core tier one capital hit 13%, having closed 2025 at 12.4%, while total capital reached 16.4% following May’s £250m additional tier one issue. Deposits rose to £18.8bn at a lower funding cost of 3.8%.

Not everything moved the right way. Impairments climbed to £50.7m from £32.6m, reflecting further provisions taken against a small legacy book of development finance, and return on tangible equity slipped marginally to 18.1%. Full-year guidance was reiterated: lending near £21bn and cost-income below 38%.

Looking Forward

Shawbrook is the second UK lender in recent weeks to tie AI to a disclosed cost measure rather than a stated ambition, following Lloyds’ half-year comments on AI-driven cost reduction. That matters for investors trying to work out which AI spending is producing anything. Its first ordinary dividend, due in 2027 and based on this year’s earnings, gives a reasonable checkpoint for whether the compounding Castrillo describes shows up in the numbers.