DON'Ts for corporate innovation

I have sat in rooms that smelled of Sharpie and optimism. I have watched a listed bank, an energy retailer, a telco, a Magic Circle alliance firm, a hospital trust, and a US card issuer all run the same week. Same icebreaker. Same journey map. Same Friday applause. Same Monday silence.

Corporate innovation has a branding problem. The logo is a lightbulb. The artefact is a mural. The outcome is a slide. We spent a decade teaching large organisations in finance, government, legal, healthcare, and energy to adopt the language of startups without adopting the constraints that make startups honest. Startups die. Your lab does not. That is the whole difference.

This is the anti-playbook. Not because I enjoy the word "don't". Because the don'ts are the only part most executive teams have not written down. The dos are already in the vendor deck. The don'ts are what protect a quarter.

The lightbulb industrial complex

If you run innovation at ANZ, Telstra, AGL, DBS, Allens, the NHS, or Capital One, you already know the pattern. A centre of excellence stands up. A wall goes teal. A cohort of "innovation champions" gets a two-day course and a Slack channel. The CIO or the CCO or the new Head of Transformation keynotes the town hall. Then the work of changing a product, a process, or a P&L line is politely returned to the people who were busy before the offsite.

I am not against workshops. Collective Campus runs them. I am against workshops that leave no residue: no prototype, no decision, no next experiment, no owner who was in the room. Residue is the only honest unit of progress.

The companies that look like they "get it" from the outside often have the same internals as the ones that do not. DBS is routinely held up as the digital bank that left the pack. Fair. The useful part is not the award reel. The useful part is that digital work at DBS was treated as a P&L problem with owners, not a mural problem with facilitators. ANZ has spent years on agile and design-led work across New Zealand and Australia. The useful part is not the framework badge. The useful part is when a squad can kill a feature without a steering pack that restates the workshop.

Telstra has run digital and AI transformation programmes that would fill a conference. AGL has had to invent inside a listed energy business that cannot "move fast and break things" because the thing you break is a household bill or a generator. Allens, through the Linklaters alliance, has the same client-intake seams every top-tier firm pretends are a knowledge-management problem. The NHS has more service-design talent per square metre than most consultancies, and still loses years to business cases that pretend certainty. Capital One built a genuine design org in the US and still had to fight the gravitational pull of the slide.

Different logos. Same physics.

If every idea graduates, none of them were experiments.

Don't launch a lab with no killing mechanism

A lab that cannot kill is a marketing department with beanbags. Write the kill criteria before the concept. Write them in language a risk lead and a CFO can both sign. "We will stop if we cannot get 200 target customers to complete the concierge version in 30 days." "We will stop if the unit economics require a subsidy the board will not name." "We will stop if legal will not let us run the paper version."

ANZ-style programmes that survive do this in public, inside the team. They do not hide the kill in a stage-gate that only fires after a million dollars. Capital One's better design work treated evidence as a reason to stop, not only a reason to scale. The NHS pattern that fails is the pilot that never expires. The AGL pattern that works is the offer test a retail and risk lead can live with, then kill without a press release.

A practical close: put the kill date on the same page as the hypothesis. Circulate it the day of the workshop. If you cannot name the kill, you do not have a bet. You have a hope.

Don't train 400 people and change zero incentives

Design thinking dies in a KPI that still rewards last year's product. I have watched telco and bank programmes (Telstra-scale enablement, DBS-scale digital academies, ANZ-scale agile coaching) graduate hundreds of people who return to a bonus line that punishes the experiment.

If the product owner is still measured on feature output, they will mural on Thursday and ship the roadmap on Friday. If the partner at Allens is still measured only on hours, they will nod at service design and staff the matter the old way. If the NHS service owner is still measured on business-case theatre, they will commission another discovery that cannot change the service.

Training is not a culture programme. Training is a prototype of a new operating rhythm. The rhythm has to pay. Change one incentive on one squad before you buy another cohort.

A practical close: pick one team. Change one metric for one quarter. Pair the training with that metric. Measure hours returned or cycle time or kill-rate, not attendance.

Don't run a hackathon to "build culture"

Culture is what you do on a wet Tuesday. A Saturday trophy does not change a Tuesday. Run a hackathon to produce three prototypes you will fund or kill in 30 days. Book the follow-up before you book the venue. Put a dollar figure on the two ideas you will actually run. Invite the person who can say no.

Capital One and Atlassian-style internal hack patterns (ShipIt, product challenges) only matter when a product owner is in the room and a path to the backlog exists. AGL-style energy hacks fail when the winning idea needs a production integration on day one and nobody from risk attended. Allens-style legal hacks fail when knowledge and intake are treated as a poster, not a journey a partner will fund.

I have written this so many times I can say it in my sleep: if nobody owns the follow-up budget, you ran a party.

Don't buy an AI strategy off a vendor deck

Start with one workflow a team already hates. Measure hours returned. Telstra, DBS, HSBC Hong Kong, and Canva all have public AI programmes. The ones that leave a residue start in a calendar, not a capability map. Where do senior people still paste, summarise, reformat, and chase? That is the value chain.

A Chief AI Officer who funds a slide titled Transformation has bought theatre with a new job title. A CAIO who funds a two-day AI-for-teams sprint on a hated workflow, with a domain lead and a builder in the same room, has bought an operating system. Same budget class. Different Monday.

Don't separate the lawyers, the risk leads, and the operators

They are users. Interview them. Prototype the paper trail. Time-box disagreement. Design thinking that cannot survive a one-page risk note was a sketch, not a bet.

Allens and Linklaters partners already run a portfolio. They feel client seams in intake, scoping, and knowledge handoff. If you run a workshop without a partner and an operator in the same room, you will get a mural the practice will not fund. ANZ risk and NHS Information Governance are not villains. They are constraints. Constraints are design inputs. Treat them as such, or keep running theatre.

The GDS pattern in the UK (show the thing, write it down, make it usable) works inside banks and energy companies for the same reason it works in government: it forces the constraint into the brief. Services Australia and NSW Government programmes that stall usually stalled because the constraint arrived in week nine as a surprise.

Don't confuse an offsite with a strategy

A strategy is a set of bets with owners, kill dates, and a resource shift. An offsite is a calendar invite. I have facilitated offsites that produced both. The difference was never the venue. The difference was whether a decision-maker sat in the room and left with a sentence they would repeat in the board pack.

If your innovation strategy cannot be said in four lines, it will be said in forty slides. Forty slides is how theatre hides.

The few moves that actually move a P&L

Here is the short list I will keep publishing until it is boring.

Frame the constraint as part of the brief. Regulation, procurement, industrial relations, clinical safety, partner leverage: write them on day one.

Prototype the decision, not just the interface. A clickable demo that nobody can approve is still a sketch. Prototype the risk note, the pricing one-pager, the ops runbook, the legal memo.

Show the board the kill criteria before the concept. Boards relax when they can see the stop. They tense when they can only see the dream.

Put a 30-day experiment budget on the calendar before the workshop. Dollars and a named owner. Not a "we will find funding" slide.

Measure residue. Prototypes funded or killed. Hours returned. Cycle time. One customer behaviour that moved. Not mural photos. Not NPS of the facilitator.

DBS-style digital work, Capital One-style design orgs, GDS-style service standards, and the better ANZ squads all rhyme. They make it expensive to pretend. They make it cheap to stop.

A one-month pause on strategy decks

Print this. Take it to the next steering committee.

1. Name the hated workflow or the customer seam. One sentence.
2. Name the constraint that usually kills the idea. One sentence.
3. Name the thinnest test that constraint will allow. Concierge, paper, shadow process. Not a platform.
4. Name the owner who was in the room. Not a centre of excellence.
5. Name the kill date and the dollar figure. Thirty days is a default, not a religion.
6. Name the metric. Hours, conversion, complaints, cycle time. One number.
7. Circulate the residue the same day. A page, not a pack.

If you do that at ANZ or Telstra or AGL or DBS or Allens or an NHS trust or a Capital One-style product org, you will have done more than most labs do in a year. If you cannot do that, do not book the venue.

Collective Campus exists to make the opposite of theatre true: short, sharp T&E that leaves a residue. A prototype. A decision. A next experiment. Not a mural.

A Monday scorecard the lab cannot hide from

Innovation labs love a quarterly pack. A Monday scorecard is meaner and more useful. It fits on one page. It is filled by the named owner of each bet, not by a centre of excellence that watched from the mezzanine. If the owner was not in last week's room, their name does not appear. That single rule retires half the theatre at ANZ-scale banking, Telstra-scale telco, AGL-scale energy, and DBS-scale digital programmes before anyone argues about culture.

Print five columns. Use them every Monday for a quarter. Then decide whether the lab is a factory or a hobby.

1. Bet. One sentence a customer or a frontline officer would recognise. Not "AI-enabled hardship." "We will test a plain-language bill explanation in one retail bay."
2. Owner. Role and name. A GM of Retail, a Head of Care, a partner at Allens, a service owner at Services Australia. "The programme" is not an owner.
3. Residue last week. Prototype funded, prototype killed, hours returned, cycle time moved, or explicitly blocked. Blocked is a result. A mural photo is not.
4. Kill date. A calendar day. If the date passed and nobody wrote a funeral note, the bet is a zombie. Zombies are how labs lose CFOs.
5. Money still live. Dollars sitting in a cost centre, not a slide titled Pipeline. If the figure is zero and the bet is still "live," you are running a hope.

ANZ squads can hang this on an existing heartbeat. Telstra care and field leads can hang it on an ops huddle. AGL retail and risk can hang it on a product committee. DBS-style digital owners can hang it on a P&L conversation they already know how to have. NHS service owners and GDS-shaped teams can hang it on a 90-day card. Capital One-style product orgs can hang it on a design-week readout. If you need a new steering pack to talk about the scorecard, you have added a room. Delete one first.

A practical: send the blank scorecard on Friday afternoon. Demand it back by 10am Monday. People who cannot fill it did not have a bet. They had a narrative. Narratives do not get another facilitator.

How a CFO reads residue

A Chief Financial Officer at NAB, Lloyds, RBC, or a Telstra-scale telco does not hate innovation. They hate open-ended spend with no stop and no unit of progress they can audit. They have seen the lightbulb industrial complex. They have signed the catering. They have watched last year's lab become this year's "capability." They will fund a test they can kill. They will starve a story they cannot measure.

Residue, to a CFO, is not a vibe. It is four artefacts:

A framed problem with a clock. Hours, cycle time, complaints, completion, exceptions. One number a controller can put next to a baseline. "We will know in 30 days" is a sentence they can live with. "We are building a movement" is not.

A kill they can see. Expiry is a control. AGL-shaped energy and NAB-shaped banking already live inside licences that punish uncontrolled change. Show the stop on the same page as the ask. Boards relax. CFOs especially.

A named owner who already has a cost centre. If the owner is a lab director with no P&L, the CFO hears "this will migrate later." Later is how labs become furniture. Put a line-owner on the page. Retail. Care. Credit. A practice group. A cluster in NSW. A director at IRD NZ.

A list of what you will stop doing if the bet works. Capacity is not created by a workshop. It is created by deletion. If the lab cannot name a process, a report, or a room it will retire, the CFO should assume the lab is additive. Additive is a tax.

I will not invent a return figure for any of those companies. I will not quote a living CFO. I will say what they already do in every other portfolio: ask for a stop, an owner, a metric, and a date. Innovation does not get a special exemption because the walls are teal.

When a lab at ANZ, Telstra, AGL, DBS, Allens, or Capital One cannot produce those four artefacts, the honest move is not another offsite. The honest move is to pause the budget until a Monday scorecard exists. Collective Campus will keep saying this until it is boring. CFOs already believed it.

What to cut from next quarter's lab budget

Labs accumulate subscriptions the way programmes accumulate rooms. Before you renew anything, run this cut list. It travels across AU, NZ, HK, SG, the US, the UK, and Canada.

Cut the champion network that has no metric. A Slack channel and a badge after a two-day course is not a network. It is a mailing list. If last quarter's champions cannot name a funded test they own, do not buy another cohort. Change one incentive on one squad instead. Design Thinking T&E dies when the bonus still pays last year's product. So does Agile. So does AI Training.

Cut the vendor day that does not sit on a hated workflow. Model bake-offs, platform tours, and "innovation showcases" are how slideware wins. Telstra, DBS, HSBC Hong Kong, and Canva all have public AI programmes. The useful ones start in a calendar. If the vendor cannot name the desk whose Tuesday will change, they are selling a keynote.

Cut the hackathon that has no Monday budget. See the hackathon piece. If nobody owns two 30-day tests before you book the venue, you are buying a party. Inkeros-style sprints and Atlassian ShipIt only pay when a path into a backlog exists. AGL-style energy weekends fail when the winning idea needs production on day one and risk did not attend.

Cut the mural refresh. Journey maps that cannot be read in a stand-up will not be used in a stand-up. Woolworths-shaped, Singtel-shaped, RBC-shaped, and NHS-shaped operators already know the seams. Pay for a punch list and a 21-day look-again. Do not pay for another 36-inch PDF.

Cut the second capability map. A new CAIO, CHRO, or Head of L&D in the first 30 to 120 days will be offered a map. One is a brief. Two is a delay. Fund a two-day AI-for-teams sprint on one workflow instead.

Cut any tool whose only user is the lab. Idea portals, innovation accounting suites, and "culture dashboards" that nobody on the line opens are souvenirs. If the operator will not log in, you bought a priest.

What you keep is cheaper and meaner: a framed problem, a thinnest test, a named owner, a kill date, a small envelope, and a scorecard. That is Lean Startup inside a regulated business. That is Design Thinking that survives the board pack. That is Productivity after the meeting. The lab that cannot live on that list should not get a next quarter.

A practical cut note for the sponsor:

We will not renew the champion programme, the vendor showcase, or the idea portal. We will fund three 30-day tests with named owners and kill dates. We will report residue on a Monday scorecard. We will delete one standing room to pay for the practice time. If that is too small to be a lab, good. Labs were the problem.