Frameworks for many teams

One team can agree how to work in a room. Fifty teams building one product can’t, and that’s the gap scaling frameworks try to fill. There are several, they disagree with each other, and most organisations end up running a variation of one. The useful skill is reading each one’s claims, and its critics’, for what they are.

Each in its own words

• SAFe (Scaled Agile Framework) groups teams into an Agile Release Train of “50 -125 people” who plan together in PI Planning, held “every 8-12 weeks”. It is the most prescriptive of these, with roles, events and configurations at several levels. Its current versions are called “Core SAFe” and, added in June 2026, “AI-Native SAFe”.
• LeSS (Large-Scale Scrum) keeps one Product Owner and one Sprint for “Up to eight teams (of eight people each)”, and adds as little as it can.
• Scrum@Scale, from Jeff Sutherland, scales Scrum with Scrum itself and aims for a “minimum viable bureaucracy”. Its blunt line: “If an organization cannot Scrum, it cannot scale.”

SAFe’s site claims results like “30 – 75% faster time-to-market”. The page gives no sample, baseline or method, so treat it as a vendor’s claim, not a measurement. We link to SAFe’s own pages rather than reproduce its diagrams, which its terms restrict.

The critics, dated

The best-known critiques of SAFe are from 2013 and 2014, written about the SAFe of that time, years before its current versions. Ron Jeffries is the fairest single voice: if an organisation “actually did what it said, they would likely prosper”, but “SAFe isn’t really Agile in its heart”. You’ll also see Martin Fowler’s 2018 “Agile Industrial Complex” talk cited as a SAFe critique. It never names SAFe, so don’t use it that way.

The “Spotify model” was a snapshot

In 2012 Henrik Kniberg and Anders Ivarsson described how Spotify organised its teams into squads, tribes, chapters and guilds. Their paper called it “a journey in progress, not a journey completed”. Many companies copied it as a framework anyway. In 2020 Jeremiah Lee, who worked there, described how it played out, including a former Spotify coach’s line that “Even at the time we wrote it, we weren’t doing it” (quoted second-hand in Lee’s article). Kniberg himself said in 2015 that it “wasn’t actually intended to be a generic framework”. He also said he had “yet to see a case where a company ended up in a worse position” after borrowing from it, so the story isn’t simply a failure either.