Working notes on the integrations operations actually run on — WhatsApp Business API, Google Workspace, Razorpay — and on how we scope, measure and hand over. Written by the engineers doing the work, not a content team.
The UK Cabinet Office tells its own buyers: fixed price only works with fixed scope. Almost no genuinely new software has fixed scope. Here is what that means for how you should actually pay.
Read the piece →
The US federal government spends $754 million a year maintaining just eleven of its oldest systems, some running languages taught to almost nobody currently in the workforce. Here is how to decide what to do about the equivalent system in your own organisation.
Read the piece →
The UK has mandated algorithmic transparency records across central government since February 2024. Most technology vendors selling into the public sector have not read the standard they will be asked to comply with.
Read the piece →
These three words get used interchangeably in planning meetings and they describe very different risk profiles. Confusing them is how a two-month refactor quietly becomes an eighteen-month rewrite.
Read the piece →
Eleven of the US federal government's oldest IT systems cost $754 million a year to maintain. None of that figure appears on a single line labelled 'cost of not modernising' — it is scattered across budgets that make delay look free.
Read the piece →
The application layer gets the planning attention. The data underneath it is where modernisation projects actually run aground, usually because nobody looked closely at it until the migration was already underway.
Read the piece →
Most AI RFPs describe a technology. The ones that produce good outcomes describe a problem, a baseline, and how success will be measured. Here is the structure that gets you real answers.
Read the piece →
Many vendor websites still say high-risk obligations landed on 2 August 2026. They didn't — the AI Omnibus pushed that date to December 2027. Here is what actually applies, and when.
Read the piece →
The DPDP Rules were notified on 14 November 2025 with an eighteen-month phased window. The era of treating India's data law as 'an Act awaiting rules' is over — here is what technology buyers need to do before May 2027.
Read the piece →
UK central government awards just 10% of its direct spend to SMEs, down from 13% in 2020. Across the EU, the average number of bidders per contract has almost halved. Here is how a smaller vendor competes anyway.
Read the piece →
The systematic evidence says the choice between building a team, hiring an agency, or going offshore matters less than how the relationship is governed. But each option still carries a different, predictable set of trade-offs worth naming honestly.
Read the piece →
Most bad vendor relationships were visible in the proposal, before anyone signed anything. Here is what to look for while you still have the leverage to walk away.
Read the piece →
A UK regulator spent two years investigating cloud pricing and found that less than 1% of customers switch provider each year. That statistic describes what lock-in looks like once it has already happened. Here is how to spot it before it does.
Read the piece →
Discovery is often sold with a claim that traces to an internal IBM training programme with no published dataset. That does not mean discovery is worthless — it means the pitch for it has been resting on the wrong evidence.
Read the piece →
The global average cost of a data breach is now $4.99 million. Most enterprise software projects treat security as a checklist at the end rather than a design constraint from the start — and the difference shows up in exactly that number.
Read the piece →
A UK regulator spent two years investigating cloud pricing and concluded competition 'is not working well.' Fewer than 1% of customers switch provider a year. Here is what that means for your bill.
Read the piece →
CIOs estimate technical debt at 20 to 40% of their entire technology estate's value. Developers rank it their single biggest workplace frustration. Here is how to size it in your own codebase.
Read the piece →
A build quote is one number in a five-year calculation. Here is what the other four years actually contain, and how to price them before you sign anything.
Read the piece →
UK government data shows AI use ranges from 82% at large firms down to 40% at sole traders — the same size gradient the US Census finds. Here is where automation pays for a business without an enterprise budget.
Read the piece →
One study anchors almost everything credible we know about AI in customer service: 5,179 agents, a 14% average gain, and 34% for the newest staff. Here is what it found, and what remains genuinely unknown.
Read the piece →
The UK government ran a rigorous, 1,000-person trial of AI coding assistants and published the uncomfortable result alongside the good one. That is what a trustworthy pilot looks like.
Read the piece →
Every ROI comparison you'll find between AI agents and RPA was commissioned by a vendor selling one of them. Here is what's actually known, and what isn't.
Read the piece →
Gartner calls it 'agent washing': rebranding an existing chatbot or automation script as an agent, with none of the underlying capability. Here is how to tell the difference before you pay for it.
Read the piece →
Nine in ten companies say they use AI. Gartner estimates only about 130 of the thousands of vendors selling 'agents' have built anything that deserves the name. This is the guide to telling the difference.
Read the piece →
Nearly a third of organisations have now declined to buy software because they could build it in-house with AI coding tools. That changes the calculation — but not in the direction most people assume.
Read the piece →
A low quote is not a discount. It is usually a different scope, wearing the same name — and the difference arrives later, as a change request.
Read the piece →
The statistic everyone quotes about outsourcing failure rates comes from a 2003 Gartner prediction with no published methodology. Here is what the peer-reviewed evidence actually says about picking a technology partner.
Read the piece →
Across five UK digital programmes, supplier arrangements contributed to at least 29 years of delay and £3 billion in cost increases. The auditors were clear about why — and the fix is in how the contract is written, not who signs it.
Read the piece →
Most of what determines whether a software contract goes well is decided before anyone writes a line of code. These are the questions that actually predict it.
Read the piece →
The GAO could rank the US federal government's legacy systems by risk because it actually looked — at age, language, hardware support and known vulnerabilities. Most organisations modernise the system that is loudest, not the one that is riskiest.
Read the piece →
Named after a vine that gradually envelops and eventually replaces its host tree, this pattern lets a new system take over a legacy one piece by piece — so a single bad cutover night can never take down the whole business.
Read the piece →
Five UK digital programmes lost at least 29 years and £3 billion to supplier arrangements. The National Audit Office's diagnosis was not incompetence. It was contracts built to buy hours instead of outcomes.
Read the piece →
A vendor telling you they are 'SOC 2 certified' has already told you something worth noticing — SOC 2 is not a certification, and the distinction is not pedantry.
Read the piece →
Most organisations adopting an AI tool never actually ask where the data goes after it leaves their system. The answer is usually available. Almost nobody asks for it before signing.
Read the piece →
An organisation operating across the EU, India and building with AI is subject to three regulatory regimes simultaneously, each with different definitions and different triggers. Treating them as three separate checklists wastes effort that a single map would save.
Read the piece →
When a regulator, a customer or your own compliance team eventually asks why an AI system made a specific decision, the honest answer is often that nobody can fully reconstruct it. That is a design failure, not an inevitability.
Read the piece →
A growing share of breaches now originate through a vendor rather than a direct attack. Your own security controls can be excellent and still not protect you, if the weakest link is a supplier you never properly vetted.
Read the piece →
Deloitte's research found mid-sized businesses report meaningfully lower returns from digital investment than large enterprises. The gap is not about ambition. It is about scoping a programme to the resources actually available.
Read the piece →
The claim that '70% of digital transformations fail' is repeated constantly and traceable to no credible primary source. The real evidence points somewhere more specific and more fixable: most transformation budgets go entirely to technology, and almost none to redesigning how people actually work.
Read the piece →
Organisations reporting the strongest bottom-line impact from AI were 2.8 times more likely to have redesigned their workflows than those reporting weaker impact. Almost every transformation budget still goes to the tool rather than to this step.
Read the piece →
A successful pilot gets celebrated, funded for wider rollout, and then quietly underperforms at scale. McKinsey's manufacturing research has a name for this pattern: pilot purgatory. Here is what actually causes it.
Read the piece →
Off-the-shelf software is cheaper right up until the point where making it fit your actual process costs more than building something that fit from the start. Most organisations cross that line without noticing.
Read the piece →
Login counts and adoption rates measure whether people opened the tool. They do not measure whether anything got better. Most organisations only track the first kind of number, and then wonder why the return on investment is hard to demonstrate.
Read the piece →
The popular change-management frameworks are frequently cited as settled science. Most of the specific statistics behind them come from the vendors selling the framework — which doesn't make them useless, but does mean they should be read as marketing-adjacent rather than independently verified.
Read the piece →
Heavy customisation of a commercial platform is the worst of both worlds: you pay licence fees like a buyer and maintenance costs like a builder, while owning neither the product roadmap nor your own codebase.
Read the piece →
Nearly a third of organisations have now declined to buy a software product because they could build the functionality in-house using AI coding tools. That is a real, measured shift — and it changes less about the decision than the headline suggests.
Read the piece →
Every tool in your stack has to talk to the others, and each connection needs maintaining as both ends change independently. A decision that looks cheap in isolation — one more tool — can be expensive in aggregate, and almost nobody counts the aggregate.
Read the piece →
Every "$50,000 to $250,000" price range you have read comes from a software agency's marketing page. None of them publish a sample, a method, or a source. Here is how to work out the number yourself.
Read the piece →
Everyone in this industry quotes the same statistic. Some version of most software projects fail, and the average one runs wildly over budget. It is repeated in board papers, vendor decks and procurement business cases, usually traced back to the Standish Group's CHAOS report.
Read the piece →
Nine in ten organisations now use AI — yet only about 6% report significant earnings impact from it, and while 64% of OECD countries run generative AI, fewer than a third measure whether it worked. Here is what separates the two groups, and how Kaizen Spark Tech builds for the second one.
Read the piece →
Treated as a real platform — version control, error handling, quota awareness — Apps Script carries more production load than people expect. And where it stops.
Read the piece →
Architecture notes, environment setup, runbooks and secret rotation. The test: can an engineer who never worked on the project operate it without a meeting?
Read the piece →
One bank credit, hundreds of orders, fees and refunds inside it. How to reconcile Razorpay settlements automatically and surface the exceptions.
Read the piece →
Automation multiplies whatever it is pointed at. Three situations where we tell clients to fix the process first and quote for less work than they asked for.
Read the piece →
Per-message pricing, template categories, BSP markup and the line items most WhatsApp quotes leave out. Written for people budgeting a build, not buying a demo.
Read the piece →Everything here is written and published through the client portal rather than hand-coded into the site, so what appears on the blog is whatever has actually been published — never a card that outlived the piece it pointed at.
Read the blogNo newsletter software yet. Email us and we will add you to the list we send new pieces to.
