‘That’s not like you,’ Silas says, glancing at his young colleague.

They are in the CID corner of the open-plan Parade Room at Gablecross. Strover is an expert when it comes to computers and a champion of all things digital, constantly berating Silas for not embracing social media.

‘And stop calling me “sir”, will you? “Guv” or “boss” but not “sir”. Makes me feel like a schoolteacher. And I hated school.’

Silas recruited Strover to work with him eighteen months ago. Apart from still calling him ‘sir’, she’s a quick learner, proving invaluable last year when they nailed a serial killer. He likes her sense of humour too, now that she’s grown confident enough in his company to speak her mind.

‘I’m a big fan of deep learning, don’t get me wrong – boss,’ Strover says, emphasising the last word in her strong Bristol accent. ‘I just don’t think computers will ever properly understand human faces.’

‘You’re in danger of sounding like me,’ Silas says. He’s never heard her badmouth technology before.

‘Airport scanners are one thing – nice, clean, head-on face shots, a constrained environment – but big unruly crowds…?’ she says. He nods, encouraging her surprising bout of heresy. ‘Forget it. Poor lighting, funny angles, grainy resolution, scarves, beards, you name it. The messy real world that the software has to work with is just too variable compared with the neat and tidy images it was trained on. I mean, take a look at this.’

She passes Silas a printout of a mugshot. ‘That’s meant to be a match for our barman in the Bluebell.’ She pauses, letting Silas study the image. ‘My cat’s arse looks more like him.’

Silas puts the mugshot down and sits back, smiling. She’s right about facial-recognition software, confirming his own worst Luddite fears. The system has tried and failed miserably to match the CCTV image of the barman with one of 21 million images of faces and distinguishing marks currently stored on the UK’s custody image database. And that’s after Strover narrowed the search first with the relevant metadata. He doesn’t know why they even bothered. When South Wales Police tested new facial-recognition software on crowds in Cardiff a couple of years back, 92 per cent of the matches were false. No wonder the Met, his old force, have abandoned using it at the Notting Hill Carnival.

‘I’ve been asking the boss for weeks if he’ll consider giving the super-recogniser unit another go,’ he says.

‘And?’

‘Not unless Kate comes back. And we know the answer to that.’