Insights
What I've been reading.
Every past pick from the Insights page's reading panel, grouped by the week it was featured, newest to oldest. Back to Insights.
Week of 5 October 2026
Difficult to Lead. Impossible to Drive.
Phil at his most unguarded. He traces a lifelong 'difficult to lead, impossible to drive' streak back through decades of work, through a family link to Charlotte Brontë's Shirley, right up to how AI has handed him more latitude than ever to just go find the problem and fix it. It's not a comfortable piece to write about yourself, and he knows it. Read alongside Ian's piece this week, it's a useful counterpoint: Phil isn't the lone-wolf key-person risk Ian's warning about, he's autonomy exercised inside a team, with sponsorship and real guardrails around him. Working with Phil really is like a box of chocolates: you don't always know what you're going to get, but it'll make you smile.Phil — microlab
A personal essay tracing a decades-long "difficult to lead, impossible to drive" pattern through Phil's career, anchored by a real family connection to the Yorke family in Charlotte Brontë's *Shirley*. He works through what that temperament has cost and earned him, how AI has expanded the scope of what one determined engineer can do inside an organisation, and what he owes the people managing him in return for the latitude he needs.TCS — TCS Investor Relations
TCS's Q2 numbers are a useful reality check on how much of the AI story in IT services is genuinely new revenue versus relabelled existing work. Annualised AI services revenue has crossed US$3.1 billion, over 10% of the book, but overall sequential constant-currency growth is just 0.5%, with operating margin steady at 24%. Both readings are true at once: AI is becoming a real line item, and it isn't yet moving the overall growth needle. Worth watching whether that gap closes or widens from here.Richard Millington — Indispensable Consulting
The strongest read this week. Millington examines what happens when consulting firms stop billing primarily for deliverables and start sharing responsibility for the outcome itself, proposing a baseline fee plus a performance fee plus a stretch fee as the commercial structure. It's a genuinely different way to price expertise, and the uncomfortable part is the one most pieces like this skip: taking responsibility for outcomes means taking on risks clients have always carried themselves. Anyone thinking about how professional-services pricing evolves should read this one properly, not skim it.Ian Johnson — Mio
Ian's own piece, and it reads like someone who's watched this exact pattern play out across a dozen mid-market businesses. The solo AI operator quietly burning through years of IT backlog is a familiar story by now: genuinely useful, genuinely risky the moment that one person goes on leave, changes jobs, or gets poached. His framing is sharp: the business hasn't solved its software bottleneck, it's traded a slow one for an existential single point of failure. Lucca is Mio's answer, giving that individual speed a governed, documented home it can survive without.Lilia Guan — ARN
A sober look at three pressures landing on Australian businesses at once: accelerating AI adoption, an intensifying cyber security threat environment, and a skills shortage that shows no sign of easing. None of these is new individually, but the piece makes a reasonable case that it's the combination, not any one factor alone, that's actually squeezing IT leaders right now.Lilia Guan — ARN
A good ANZ example of a trend worth watching across the sector: a technology implementation specialist acquiring a 40-person strategy consultancy to move further into business transformation and advisory work. It's the same logic showing up repeatedly lately, that pure technical delivery capability is worth less on its own than when it's paired with people who understand how a customer actually needs to change the way it operates.Robert Young — Numbers & Judgment
This week's left-field pick, and worth reading for the question it raises more than the deal itself: a £2.24 billion acquisition that concentrates Informa's portfolio around fewer, stronger assets. What actually makes a specialist business valuable: recurring customer relationships, ownership of market access, focused operations, and the ability to scale without the cost base scaling in lockstep. Good discipline to borrow when thinking about any business, not just events and exhibitions.Lilia Guan — ARN
Ingram Micro building out a formal partner alliance structure, aimed at giving channel partners a clearer, more coordinated way to work together rather than each negotiating the relationship separately. Useful context for anyone watching how distributors are repositioning themselves as the channel itself keeps consolidating.
Week of 28 September 2026
Prathik Jayaprakash et al. — Reuters
Accenture's results did more to calm the "AI replaces consulting" narrative than a year of commentary has managed, shares up 22% on a forecast that beat expectations and dragged the rest of the sector up with it (Cognizant, IBM, Wipro and Infosys all rallied). The number worth sitting with is buried in the good news though: pricing was lower in many areas because clients are demanding a share of the savings AI delivers. Demand isn't disappearing. Margin is being renegotiated.Atul Mehta — EY
AI-driven cost savings can mask weakening demand just as easily as they can signal real structural improvement, and the only way to tell the difference is to look past the headline growth number into where the earnings are actually coming from. Due diligence has to evolve the same way the sector is, or an AI story becomes a convenient cover for a demand problem rather than proof there isn't one.Conversational Engineering
My mate Phil built this one himself, a real instrumented experiment in working collegially with frontier AI, not just another AI take. He doesn't usually put his own work out there like this, so credit where it's due. Good on you, Phil.Phil — microlab
A genuine experiment rather than a think-piece: Phil built microlab to work collegially with frontier AI on a real, instrumented system, treating the collaboration itself as something worth measuring rather than assuming. Early days, but it's the kind of first-hand build-and-measure work that's rarer than the commentary written about it.James Bergin — Xero
The "harness" framing is the useful bit here: rather than asking every small business to become its own AI governance expert, the argument is that the platform itself should embed the permissions and validation checkpoints. Ties directly into the guardrails-versus-governance distinction from here a few weeks back, just argued from the platform-vendor side rather than the user side.Lean LaunchPad - The Next Generation
From Rory, our wise owl on the advisory board. Steve Blank doesn't just lament what AI broke in his teaching, he's rebuilt the whole curriculum around it: Initial Untested Products instead of MVPs, Design Partners as real validation evidence, AI-specific risk categories built into the business model canvas itself. Worth reading as the methodology catching up to the problem he flagged a fortnight ago.Steve Blank — Steve Blank
Blank doesn't stop at diagnosing the problem the way his earlier piece did, here he rebuilds the Lean LaunchPad curriculum around it. Teams now build "Initial Untested Products" rather than MVPs, use "Design Partners" as real validation evidence, and the business model canvas itself gets new categories for LLM cost, model portability risk and autonomy permissions. The course now spends its first weeks on deep problem and stakeholder discovery before any building starts, a direct answer to AI making it too easy to skip straight to a polished-looking product.Pierre Briand — SCD Advisory
Two deals inside this monthly roundup make the point better than the headline does. Alceon buying a Microsoft Copilot specialist, and Netwealth paying over $20m for an AI advice-automation platform, both say the same thing: acquirers are paying for narrow, applied AI capability bolted onto a real client base, not for scale on its own. Worth the scan even if M&A isn't your thing, it's a live read on what capability actually trades for right now.Lilia Guan — ARN
Another roll-up, and the pattern across Evergreen's last several acquisitions is consistent: buy into security-sensitive verticals (healthcare here), keep the acquired leadership and brand, and bet that AI-enabled advisory is what makes an MSP defensible rather than commoditised. Worth tracking as a live example of the consolidation thesis playing out deal by deal rather than in the abstract.Progress Software management — The Motley Fool
A long one (around 25 minutes), but worth it for the Domo integration detail: Progress is deliberately shifting that acquired business away from lower-margin professional services toward consumption-based pricing, accepting a near-term margin hit (100-200 basis points) to get there. A real-time example of a platform vendor trading short-term margin for a pricing model it thinks is more durable.Ksenia Kartamysheva — Birdview PSA
A vendor blog, but the SPI number inside it is the real story: mature firms run 81.2% billable utilisation against 54.7% for the least mature, a gap that dwarfs anything a pricing change or an AI tool will move on its own. The four-stage model (Improvised, Coordinated, Integrated, Predictive) is a reasonable way to locate where a firm actually sits before reaching for a bigger fix.Reuters — Reuters
Outside the usual IT-services orbit, but the deal structure is the interesting bit: a $4bn enterprise value built on debt, rollover equity and a staged option on the remaining 35%, rather than an all-cash exit. A reminder that consolidation financing doesn't have to mean selling out in one go, there's a real structure here for sellers wanting to keep skin in the game.
Week of 21 September 2026
8 insights from the 'Certinia at Dreamforce' event: AI resets the economics of professional services
Cheryl Knight — SiliconANGLE
Eight takeaways from Certinia's Dreamforce event on where AI is actually taking professional-services economics: fixed-bid replacing hourly as firms absorb delivery risk in exchange for compressed timelines, deterministic systems built on continuous context capture and knowledge graphs that compound intelligence across engagements. The line worth sitting with: labour demand on an individual project is falling while overall client demand keeps rising, because price points are coming down as value delivery goes up.OpenAI — OpenAI
OpenAI's own response to the incident METR investigated independently, worth reading as the paired second half rather than on its own: their account of what happened and the road ahead, set against METR's finding of what the agents actually did during it. Two different vantage points on the same real-world incident is more useful than either alone.The Year AI Came For Us: Teaching Entrepreneurship Will Never Be The Same
From Rory, our wise owl on the advisory board. AI just kicked the legs out from under how Stanford teaches entrepreneurship, students now arrive with a finished product before ever talking to a customer. Worth reading before you assume that's only a classroom problem.Steve Blank — Steve Blank
Steve Blank on AI breaking his own Lean LaunchPad teaching model: students now build a polished, complete product on day one using AI, before doing any real customer discovery. He calls it "evidence theater", a finished-looking deliverable that mimics progress while actually increasing confirmation bias and delaying the pivots that used to come from the hard, iterative work of talking to customers. His sharpest line: when everyone can build quickly, what you choose to build and for whom becomes the whole game. The competitive bottleneck has moved from engineering speed to judgement, which is exactly the skill his curriculum was built to teach and exactly what AI lets students skip.Sarah Edwards — Kantata
Firms aspire to outcome-based pricing but usually aren't actually ready for it: no real visibility into true delivery cost, no measurement of the outcomes actually achieved, no repeatable delivery process to price against. The CPO's own framing is the useful bit here, that most firms are one or two levels of operational capability away, not one bold commercial decision away. Worth reading for anyone still treating the pricing model itself as the lever, when the real gap is usually operational.NCSC — NCSC New Zealand
NZ's NCSC on AI reshaping the threat landscape: frontier models lowering the bar for automated attacks and highly personalised targeting, cybercrime severity already at record levels through 2025/26. Their own conclusion cuts against the hype a little: strong cyber security fundamentals remain the best defence against both human and AI-enabled threats, a useful reality check before reaching for an AI-specific tool as the fix.Introducing System One Models & Jev
From my mate Phil. More technical than most of what lands here, worth it anyway. A genuinely different model architecture, not another LLM wrapper.Diogo Almeida — TypeSafe AI Blog
TypeSafe's Jev borrows Kahneman's System 1/System 2 framing literally. Most LLMs are built for System 2 style output, flexible, deliberate, conversational text, which is exactly why they're unreliable for automation: they generate unpredictable strings that need parsing, and they're consistently overconfident about being right. Jev is built the other way round, producing structured, calibrated values a piece of software can act on directly, generated in parallel rather than token by token, landing in 70-500 milliseconds against 3-plus seconds for a frontier LLM. The training write-up itself is more technical than most of what lands here, but the implication doesn't need the mechanics to land: a model built to be trusted by the software calling it, not just by the person reading the chat window.MSP Process — MSP Process
MSP Process (identity verification, AI voice, Teams ticketing) buying Triggr (workflow automation built specifically for MSPs) to close the loop between verifying who's doing the work and automating the work itself. The CEO's framing is blunt and correct: verify the user, automate the work, deliver a more secure and efficient service without adding headcount. Another data point for the platform-consolidation wave running through the MSP tooling market.FuturumAI — Futurum Group
Datacom quietly building the dominant sovereign compute position in New Zealand: bought T4's Auckland data centre, now five owned-and-operated sites, $200m+ total infrastructure investment, all upgraded for AI-ready high-density workloads including liquid cooling. Sovereign here means locally-owned and governed, which matters increasingly to government and regulated enterprises wanting out from under offshore hyperscaler priorities. Worth watching who ends up capturing the consulting and managed-services revenue that follows.IBM — IBM
IBM picking up a UK cybersecurity firm specialising in Secure by Design work for defence, government and critical infrastructure, terms undisclosed. The real angle is digital sovereignty: enabling UK clients to adopt AI while keeping control over sensitive data and meeting regulatory requirements, not just another cybersecurity bolt-on.
Week of 14 September 2026
Brief independent investigation of agents' behavior, reasoning and collaboration in the OpenAI / Hugging Face hacking incident
From my mate Phil. Unusually useful evidence for agent-governance conversations - what agents actually did under real conditions during a real incident, not what a governance framework assumes they might do.METR — METR
An independent investigation into how AI agents actually behaved, reasoned and collaborated during a real security incident, not a theoretical account of how they're supposed to behave under a governance framework. Older now (late August), but exactly the kind of grounded, after-the-fact evidence that's more useful for conversations about agent oversight than most governance-framework papers doing the rounds.Consultancy.com.au — Consultancy.com.au
Particularly relevant locally: First Focus's fourth NZ acquisition this year, adding Wellington-based Resolve Technology's roughly ten people (cloud/data hosting, IT management, cybersecurity for legal, healthcare, government and NGO clients) and pushing the NZ operation past 100 specialists. The more interesting detail is what happens to the founder: Simon Falconer moves into a virtual Chief AI & Innovation Officer role across the combined business, rather than just an earnout and an exit. A concrete example of a consolidator buying differentiated capability, not just customers and recurring revenue.Judgment Layer — Judgment Layer
The most interesting Adapt piece this week, even though the newsletter packaging undersells it. The argument: legal training has always been unsupervised learning, junior staff learning by trial and error on real work, quality depending entirely on which partner you happened to be staffed with. AI lets firms make it supervised instead: senior experts define the standard, AI delivers just-in-time instruction against it, and competency stops being a matter of luck. Worth reading past the Substack presentation for the underlying claim about what happens when you can finally design how judgement gets built at scale.Niobrara Capital — Business Wire
A US private equity firm acquiring a 30-year-old pan-Canadian MSP as a platform to consolidate a market its Managing Director calls fragmented, with AI adoption creating new demand. Terms weren't disclosed, but the stated plan, using MSP Corp's national footprint and its own acquisition-integration track record to roll up further and push into the US, is the standard PE-into-MSP playbook right now. Another data point for how much consolidator capital is chasing recurring-revenue IT services businesses.Kyndryl — Kyndryl
The completion, not just the announcement, of Kyndryl's Healthcare IT Leaders deal: up to US$350m including a performance earnout, for a business that's roughly 1% of Kyndryl's FY26 revenue. Small relative to Kyndryl's size, but strategically specific, folding Healthcare IT Leaders' EHR/ERP/revenue-cycle consulting into Kyndryl's infrastructure and AI capability so hospital systems get one provider across the stack. Kyndryl also paused its share buyback to help fund it, which says something about how much conviction is behind the healthcare push.Miranda Brownlee — Accountants Daily
A newly created role at Prospa, going to Angelo Azar, twenty-plus years in financial services and digital banking operations, most recently COO at Ubank and a founding executive/COO at Honey Insurance where he scaled AI integration. CEO Greg Moshal frames it as investment in how the business grows rather than a response to a problem. Worth noting as a leadership-maturity signal: a fintech lender formalising an operating layer as it scales, not just adding headcount.Protera — Protera
Protera repositioning SAP application managed services around outcomes rather than the traditional break-fix hourly model: proactive monitoring ahead of user impact, unused hours rolling over instead of being forfeited, and a root-cause module surfacing recurring issues rather than just closing tickets. CEO Mike BeDell's line, support should be built around outcomes, not hours, is the whole pitch. Relevant to anyone watching how SAP AMS providers are trying to differentiate as RISE with SAP and Business AI Platform reshape what's actually billable.
Week of 7 September 2026
Deltek — Deltek
A personal favourite, and one I've contributed to many times over the years. The SPI benchmark is properly professional-services-specific rather than a generic software or tech-sector cut, and the value is in the long-run trend lines: billable utilisation down to a record-low 66.4%, project margins up to 37.7%, revenue per consultant at US$210k, and blended profitability still stuck at 9.9%. Now in its nineteenth year. The full report is worth the download, not just the summary.Jackson Wilson — Oliver Group
A grounded valuation primer: normalised EBITDA multiples of roughly 3.0 to 6.5 times, with genuine contracted recurring revenue supporting the top of that range and project or break-fix work dragging it down. The useful reminder is that a single sector multiple on total revenue is the wrong model; where a business lands depends on contract quality, client concentration and how much of the revenue is really recurring. Relevant to any owner thinking about an eventual exit, or just about what the business is building toward.Upwork Research Institute — Upwork
A useful counterweight to the hype. SMB leaders are confident with AI agents (62% comfortable handing over high-stakes tasks, 32% calling them mission-critical) and the returns are strong for those past the pilot stage: 93% saw revenue grow, 91% report a positive year-on-year return. But the productivity gains are described as incremental, not transformative, and the functions actually scaling are the unglamorous ones: data analytics, content generation, inventory. Worth reading for where real adoption is versus where the noise is.Bosco Tan — Weel
Real card-spend data from Australian and NZ SMBs, not survey intent. Anthropic's share of AI-adopting businesses reached 17% by June, closing the gap with OpenAI from 13 points to 2, and Anthropic customers now spend roughly 3.6 times more per month ($1,082 against $298). The telling part is that this is genuine switching rather than new adopters choosing differently: OpenAI-only subscriptions fell from 78% to 38% of the base. AI adoption crossed 30% for the first time while traditional SaaS penetration slipped, which is the longer trend worth watching.Jason Hartley — TechRadar Pro
"AI visibility" is quietly splitting into two problems. Being understood and recommended by the model is one thing; staying visible once the platform doing the recommending also sells ad placements inside the conversation is another. Amazon's Sponsored Prompts and Google's move to put advertising into conversational search are the early signs. For any business whose customers will increasingly ask an assistant rather than run a search, this is the shift from "can the model find us" to "can we still be chosen when part of the answer is paid for."Thomas John — ITPro
The argument: most organisations are using AI somewhere (88% in at least one function) but stuck at pilot stage because the AI sits beside the workflow rather than inside it. The opportunity for MSPs is the one cloud migration created, helping customers redesign processes around the new capability and then staying on to optimise it, rather than billing a one-off implementation. Familiar logic, but a clear statement of where the recurring revenue is meant to come from.Louis van Wyk — Reseller News NZ
Forty years from PC assembly in Palmerston North to a managed services provider with a team of around 70 and customers across the Tasman. Worth reading as a New Zealand case study in the arc every services business is on: the product you sell changes completely, more than once, but the business survives by staying close to the same customers through each shift. A quieter counterpoint to the acquisition-driven growth stories.
Week of 31 August 2026
Tomasz Tunguz — Tomasz Tunguz
An older piece, but the underlying data point still holds: PS margins in software vary wildly by strategy, from firms using services to win the deal to firms that treat it as a cost to minimise. Worth revisiting now that "AI-native services" is forcing the same margin question all over again, just with a different delivery mechanism.PwC Australia — PwC Australia
PwC turning finance-function work into an agentic managed service, not just an AI-assisted project, is the shift worth watching: a blended pod of specialists and agents handling AP, AR, reconciliations and close, priced and delivered as an ongoing service rather than a one-off engagement. Advise, build and operate collapsing into a single motion, aimed squarely at the mid-market PwC couldn't previously serve profitably. It's close to the exact model I keep describing for Mio, just with a Big Four balance sheet behind it.Daniel Todd — ChannelPro
A shared-license pool that lets MSPs provision SASE per customer without a separate purchase order every time is a small mechanic with a big implication: it's licensing catching up to how MSPs actually sell, in drips, not in blocks. The self-service angle matters more than the SASE angle here — it's the difference between a partner programme built for the vendor's sales cycle and one built for the reseller's.Equiteq — Equiteq
A boutique healthcare IT consultancy folding into Kyndryl's infrastructure and AI modernisation arm is a fairly clean signal of where acquisition appetite sits right now: not generic MSPs, but teams with genuine vertical depth a generalist can't build internally on a reasonable timeline.Chad Fowler — The Phoenix Architecture
Fowler's reframe is sharper than the standard "AI kills SaaS" take. The company-shaped software model survives; what doesn't survive is the assumption that ten thousand companies should all run the same interface on top of it. His pace layers idea, standardise the slow-moving foundation (identity, data, compliance) and generate the fast-moving surface (workflows, dashboards), is close to the shape I keep seeing at Mio: standardised, secure infrastructure underneath, genuine customisation on top instead of the same-same every SMB gets from off-the-shelf SaaS. Feeding directly into the joint Dutch Business Association and Irish Business Network New Zealand event in September.Vallari Srivastava — Reuters
SLB paying $4.1 billion for a data centre cooling company is the kind of deal that only makes sense once you accept AI infrastructure build-out isn't purely a software story — it's a thermodynamics story. Kelvion's data centre revenue alone is running at $1.2-1.3 billion and growing faster than the rest of its book, which tells you where the real capital is actually going: not the chips, the plumbing around them.Tomasz Tunguz — Tomasz Tunguz
The mechanics of why renewal revenue eventually outweighs new bookings past $25M ARR, and why customer success ends up carrying as much strategic weight as sales. A decade-old argument, but the same shape as the "stop chasing new logos, protect the base" conversation happening in services businesses right now.Harry Dev Singh — Consult Recruitment NZ
Useful ground truth from the actual NZ market rather than a vendor's framing of it: professional services hiring stays cautious and cost-conscious, firms are leaning harder into the shift from compliance provider to strategic adviser, and AI has moved from an experiment to a practice-management and productivity issue that's just expected now. Candidates are weighing the whole career proposition, not just salary, which raises the bar for what a "good place to work" actually has to mean.ChannelPro Team — ChannelPro
Another AI capability arriving pre-packaged for white-label resale rather than requiring MSPs to build or integrate it themselves. That's the trend worth tracking, not the call-handling feature set: vendors are increasingly doing the AI-native engineering work upstream so partners can sell the outcome without owning the complexity.Anjali Fluker — ChannelPro
The pitch isn't the platform, it's the reframing: visibility into telecom and mobility spend as a recurring service rather than a one-off audit. Finding 20%+ in overlooked savings is a good headline, but the durable point is that MSPs keep finding new categories of "boring infrastructure" to wrap a recurring-revenue service around.
Week of 24 August 2026
Belle Lin — Business Insider
KPMG's own vice chairman admits it: a decade ago they'd have called themselves a time-and-materials business with smart people doing smart things. Now the Big Four are racing to look like tech firms because that's genuinely how clients want to consume their expertise. Worth reading for the harder question buried underneath: if the deliverable gets radically cheaper to produce, where does growth actually come from next?Raja Pabba — California Management Review
The real claim here isn't about pricing. It's that expertise itself becomes software, and firms that can't encode their judgment into systems are defending revenue that's already being built around them. Worth reading if you've ever priced a project by the hour.EY India — EY
449 deals, US$14.8 billion, and a quiet rule change underneath it: buyers are no longer paying for AI potential, they're paying for AI already showing up in the numbers. If you're building toward an exit, that's the bar now, not the story.Shrikant Umrikar — Express Computer
A real, anonymised case study buried in an opinion piece: a firm billed hours to build automation, the automation removed the hours it could bill, then a fight broke out over who owns the bots because nobody wrote an ownership clause for a deliverable that isn't a person's time. The sharpest version yet of the argument this list keeps circling: change the delivery model and leave the contract untouched, and the value just leaks out the side.Michael Gerstenhaber, Pravir Gupta — Google Cloud
Agent workloads don't fit old billing models: too bursty for a flat subscription, too risky for open pay-as-you-go. Google's fix, hybrid billing plus spend caps, is a preview of the FinOps problem every business running agents at scale is about to hit.Kyndryl — Kyndryl
Kyndryl is packaging AI agents into pre-built modernization workflows, a direct answer to the now-familiar problem: most enterprises scaled GenAI without scaling the results. Worth watching if you want an early read on where infrastructure-services margins are heading.EightM — EightM
The multiple gap keeps widening: sub-$30M platforms sit around 3 to 6x EBITDA, scaled 80%+-recurring MSPs clear 8 to 12x and up. "Recurring revenue" used to be a nice-to-have on a pitch deck. It's now the whole valuation conversation.The Death of the Company
Views from the edge from my mate Phil. "If you're not living on it, you're not living." The dark side point of view that always provokes thought.Peter H. Diamandis — Metatrends
Diamandis at his most maximalist: 70% of CEOs admit two people with AI agents could replicate their best business line in 90 days, and he thinks that's the whole story. Take the specific numbers with a pinch of salt, but the underlying claim, that coordination costs have collapsed to near zero, is worth sitting with even if the timeline is nonsense.Route1 — Route1 / ACCESS Newswire
Revenue fell on lumpy device and services timing, but margin climbed to 45.2% and recurring support revenue grew 12.4%. A reminder that the headline number and the health of the business aren't always the same story.
Week of 17 August 2026
Reuters — Reuters
Clients are rewriting IT services contracts around AI-driven productivity, pushing Indian outsourcers to deliver more work for flat or lower fees rather than paying a premium for AI capability.TSIA — TSIA
Makes the case for treating services delivery as an engineering discipline rather than a series of bespoke projects - standardising the repeatable parts so margin doesn't depend entirely on the individual consultant running the engagement.Bo DiMuccio — TSIA
Six predictions for how AI Economics is reshaping tech services - the end of SaaS-era complacency, services getting refactored around outcomes rather than hours, and what adapting actually requires.Bo DiMuccio — TSIA
Older but still holds up: the operational case for moving professional services off time-and-materials and onto a subscription model, and what has to change internally to make that work.Victoria Durgin — Channel Insider
MSP buyers are paying up for recurring revenue mix and operational maturity over raw revenue size - a reminder that valuation multiples increasingly reward how a business runs, not just what it bills.Roberto Sanabria — Iconic
Breaks down the specific factors buyers are underwriting in 2026 MSP deals - recurring revenue quality, client concentration, margin durability - beyond the headline multiple.Nicole Kiriakopoulos — PCE Investment Bankers
A banker's-eye view of current deal activity and pricing across business services - useful as a sanity check against the more MSP-specific valuation pieces.ISG — Information Services Group
ISG's index data on record Q2 growth in the global tech services market, with AI-related demand cited as the primary driver - useful market-level context behind the individual deal and contract stories elsewhere in this list.