The Delivery Blog
How AI delivery actually goes inside a company: where models belong and where they do not, what discovery finds that a requirements document misses, and what keeps a system correct after it ships.
Featured Articles
Why AI Pilots Fail to Scale (and How to Get One Live)
AI pilots rarely fail on the model. They automate the documented process, leave exceptions unowned and can't prove they still work. Six causes and the fix.
AI Copilots for Construction Quoting: Why Full Automation Is the Wrong Goal for Computo Metrico
Construction quoting is high-stakes, not repetitive. Here's why an AI copilot with human review beats a fully autonomous agent on computo metrico work.
AI Automation for a COO: What to Automate First
A practical framework for operations leaders: where to start with AI automation, how to build the business case, and the failure modes that stall programs.
Recent Posts
AI Evaluation Suites: Is Your System Still Right?
A live AI workflow drifts quietly as price lists, supplier formats and the model change. What an evaluation suite is, and what to ask for so you can tell.
AI Pilots in a PE Hold Period: A Playbook
A sponsor's AI pilot has one job: a written yes or no per portfolio company before exit. Why the short PoC stalls, and what goes in the board pack.
What an AI Implementation Partner Should Hand Over
You can't inspect a client list or a pod of engineers. You can inspect a document. The five a partner should produce, and what to ask before signing.
Embedded Operator vs Staff Augmentation vs AI Pod
Three offers that all say 'embedded'. Staff augmentation sells hands, a pod sells capacity, an operator sells judgement. Who decides, and what you keep.
Build vs Buy for AI Workflow Automation
Both decks skip the question that decides the return: how much of the workflow needs a model at all? When to buy a licence, when to build on your ERP.
PDF Order Data Extraction for Manufacturers: Automating Orders That Never Look the Same
PDF order data extraction for manufacturers: turn inconsistent customer purchase orders into structured data, checked against your own certifications.
Marketing Automation Platforms Compared: HubSpot, Mailchimp, Klaviyo, Salesforce (2026)
Marketing automation platforms compared on verified 2026 pricing: HubSpot, Mailchimp, Klaviyo, and Salesforce Marketing Cloud, plus how to choose one as an SME.
Autonomous AI Workflows: Evaluate One Before You Build
A three-question test for evaluating autonomous AI workflows step by step, plus why only 6% of enterprises fully trust agents with core processes today.
Process Mapping Examples for Italian SMEs (Italian guide)
Italian guide with a worked process-mapping example: order-to-invoice swimlane, documented vs real process, and an exception ledger excerpt.
No IT Department? How to Map an Acquired Company in 5 Days
Post-merger guides inventory endpoints and licences — which answers nothing in a 50-person target. The 5-day method for capturing how the work actually runs.
Pre-Launch Testing for Custom Software: Why the Final Weeks Should Be All QA, No New Features
Custom platforms rarely fail from a UI bug. They fail when pre-launch testing skips the flows where payments, bookings, and confirmations meet real contention.
How to Measure AI Agent ROI: The Three-Bucket Framework
How to measure AI agent ROI: revenue uplift, cost saving, and risk mitigation are the only three buckets that matter. Price every agent against all three or it can't be defended.
AI Agent Audit Trail: Why It's a Trust Product, Not an Engineering Afterthought
An AI agent audit trail is what lets a client or a regulator trust an autonomous system enough to leave it running. The EU AI Act just made logging mandatory for high-risk systems — here's what a real one requires.
When Excel Breaks for Sales Reporting (And What Actually Fixes It)
Excel sales reporting doesn't fail randomly. It fails at a predictable point: the moment one spreadsheet has to answer to more than one stakeholder. Three signs it's happened, and why the fix is a schema, not a dashboard.
AI Implementation Partners: The Market Has a Missing Middle
Firms that embed engineers price for Fortune 500 budgets, and a platform alone rarely lands the change. What a mid-market firm is really choosing between.
Most of Your AI System Should Not Be AI
Gartner predicts over 40% of agentic AI projects will be cancelled by 2027, mostly for reasons that have nothing to do with model quality. The systems that survive share an unglamorous trait: most of them is not AI. A framework for deciding which steps need a model, which need code, and which need a person, before you build.
The Documented Process Is Not the Real Process
Ask what step one is and you hear "an email arrives." The reality: dozens of senders, no two alike, routing in one head. Why discovery comes first.
AI Integration With Legacy Systems: The Data Layer
AI assistants stall on legacy ERPs with no API. How to integrate AI with legacy systems through one data layer, without replacing the ERP.
Goods Receipt vs. Invoice Matching: How to Automate 3-Way Matching in Accounts Payable
Goods received but the invoice doesn't match? Automated goods-receipt matching cuts AP cost per invoice from $9.84 average to $2.65 for Best-in-Class teams (Ardent Partners 2025).
Your Leadership Team Already Disagrees. The Meeting Is Hiding It.
Executive agendas are ordered by function, so contested items arrive when the time has gone and get deferred. Why consensus in the room is often an artefact of sequence, where AI belongs in the process, and why the fix is a redesigned ritual rather than a meeting tool.
Your Project Will Not Be Killed by Engineering. It Will Be Killed by the Community.
Engineering risk gets a spreadsheet and an owner. Community acceptance risk gets an adjective. Why social opposition now moves faster than permitting, how to score it before it becomes a crisis, and why the model must never touch the arithmetic.
AI ROI Does Not Come From Cutting Headcount
Gartner surveyed 350 executives at $1B+ companies and found workforce-reduction rates were nearly equal between high-ROI and low-ROI groups. Why the cut-and-book business case fails, and what the higher-return companies measure instead.
AI Does Not Respect Your Org Chart
Processes run sideways, org charts run down, and people absorb the gap. Automate the process and that stops. Why AI stalls at the seams, and what fixes it.
Why Innovation Programmes Stall Without Operators
Stalled innovation programmes rarely lack ideas or budget. They lack someone who owns shipping. Why the last mile has no owner, and how to spot yours.
Embedded Operators: How to Buy AI Delivery That Ships
What the forward deployed engineer model is, how it differs from consultants and systems integrators, how to buy it at mid-market scale, and when not to.
AI-Native Isn't a Strategy Deck: What It Actually Changes in Your Operating Model
AI-native is losing its meaning. The testable version: if you switch the system off for a day and the team reverts to the old manual process, you added AI rather than redesigned around it. What actually has to change, and how companies get there one workflow at a time.
How to Label AI-Generated Content Under the EU AI Act (Article 50)
Article 50 of the EU AI Act applies from 2 August 2026. A practical implementation guide: building the AI output inventory, dual-layer labelling, C2PA content credentials, chatbot disclosure, and where machine-readable marking is technically feasible.
Why Corporate Innovation Matters: The Cost of Standing Still
Incumbency is no longer protection: S&P 500 tenure has fallen from ~33 years to ~15. Why corporate innovation matters, why incumbents fail, and where to start.
Pre-Mortem Workshop: Turn Risks Into Opportunities (2026)
A pre-mortem surfaces ~30% more failure causes than "what could go wrong?" Here is a two-workshop playbook that turns those failures into opportunity areas.
How Businesses Are Automating Calls, CRM, and Appointments: The Complete 2026 Guide
We spent a year building and operating an AI phone assistant for SMBs before winding it down. Here is what that taught us about how businesses actually automate calls, appointments, and CRM — what works, where the per-minute economics break, and where the ROI really is.
Choosing Low-Carbon Materials at the Design Stage: How Architects Decide Before Certified LCA Assessment
Most embodied-carbon decisions get locked in during design, yet certified LCA assessment often arrives once the project is nearly fixed. Here is how architects compare material strategies before that point — with Elementa as a concrete example.
Automating Insurance Claims in Healthcare: How Clinics and Dental Groups Reclaim the Hours Lost to Paperwork
Italian clinics, dental groups, and healthcare back offices lose hours every week on coverage checks, claim submissions, and insurer follow-ups. Here is what can actually be automated — and where to start.
AI Statistics for Italian SMEs 2026: Only 15.7% Use AI — Double 2024's 7.7% (Istat & Polimi Data)
Every 2026 statistic on AI and automation in Italian SMEs: 15.7% adoption (Istat), 76% not investing, the North-South divide, and cases with real ROI.
Last-Mile Delivery Eats 53% of Shipping Costs: 5 AI Route Optimization Case Studies (UPS, Instacart, Anheuser-Busch)
The last mile accounts for 53% of total shipping costs. Verified case studies — UPS, Instacart, Anheuser-Busch, Tesco — show how AI route optimization cuts it.
AI Patient Access Case Study 2026: $586K Average Annual ROI from Healthcare Call-Center AI Agents
AI call-center agents return $586K average annual ROI per healthcare org — and 15% of deep-EHR-integration adopters clear $1M. Five named case studies.
How Agencies Scale SEO Content Across 10+ Clients Without Hiring (2026 Playbook)
The multi-client content wall is coordination cost, not writing cost. Where agency hours really go, why hiring more writers makes margins worse, and how a systems-based content operation — workspaces, voice profiles, generated calendars, pipeline publishing — breaks through it.
Fast Construction Quotes: Why the First Response Wins the Job (and How to Answer in 24 Hours)
In private construction work, the first serious offer anchors the negotiation: responding within an hour makes qualifying the client 7x more likely, yet a manual quote costs 4-8 hours of technical office time. How to compress the takeoff → pricing → offer flow into one step, with Compivo as a concrete example.
Excel Pivot Tables Still Run Sales Reporting at Most Italian SMEs: What It Actually Takes to Automate It
Most Italian SMEs still build sales reports by hand in Excel pivot tables. Here is why the real barrier to an automated sales dashboard is the data model, not the interface.
DC Power for Data Centers: Why Hardware Makers Are Flying Blind — and the Digital Platform Play That Fixes It
The shift to 800 VDC power in AI data centers re-opens the spec of every component in the supply chain. Why hardware-first manufacturers are flying blind on hyperscaler demand — and how a simulation or procurement platform turns demand discovery into a product.
AI for Private Equity, Funds and Family Offices: The Supalabs Guide
Where AI creates real value for private equity funds and family offices: deal sourcing, due diligence, portfolio monitoring and value creation. With DOGE as the practice specialized in this segment.
Automating an Online Coach's Repetitive Work: The Supalabs Guide
How Supalabs creates value for online coaches with technology: which repetitive tasks to automate (onboarding, plans, check-ins, questionnaires) and how to reclaim hours to put back into coaching. With FitSuite as a concrete example.
AI Agents for Business: A Practical Deployment Guide for 2026
Learn how AI agents for business are being deployed in 2026: use cases, ROI data, a 4-phase playbook, and how to avoid the 5 most common failure modes.
AI Efficiency Program for Marketplace Groups: 2026
Why marketplace groups need a different AI efficiency approach — cross-vertical patterns, data-moat unlock, payment-layer AI, M&A integration as reusable asset.
AI Efficiency Program for Financial Services: 2026
How banks, insurers & asset managers run an AI efficiency program under EU AI Act, MiCA, Solvency II & DORA — 3LoD-aligned, gated, audit-ready.
AI Efficiency Program for Post-IPO Scaleups: 2026
How CFOs at newly-public scaleups scope an AI efficiency program post-IPO that survives earnings calls, SOX, and the first-year operating leverage mandate.
AI Culture Transformation: Enterprise 2026 Playbook
The parallel workstream most AI programs miss: how to actually shift enterprise culture toward AI adoption — champions, executive modelling, and KPIs that matter.
How to Choose an AI Implementation Partner (2026)
Ten questions to ask an AI implementation partner, with good and bad answers, an honest comparison of partner types, red flags and a day-30 gate.
