When an executive committee asks "how do we structure our AI approach?", there are three major answers on the market. Three frameworks, three philosophies. And honestly, I designed one of the three, so let me put the transparency upfront: I am both judge and party. I will still try to be fair.
This article compares IAgile (my framework, iterative and rooted in agility), AUGMENT by Thiga (6 steps, acculturation first), and AI Ladder by IBM (data-centric, linear). Each has its strengths. Each has blind spots. And the best choice depends on where you are, not on who sells it.
IAgile: agility applied to AI transformation
Let us start with the one I know best, obviously. IAgile starts from a simple observation: AI projects rarely fail for technical reasons. They fail because organizations over-plan, deliver too late, and the organization does not have time to digest the changes.
The framework works in short sprints (2 to 4 weeks) applied to the transformation itself — not just to individual projects. Use case identification, rapid prototyping, field testing, feedback loop, adjustment. The same iterative cycle as Scrum, but applied to the overall AI strategy.
The key idea: you do not define 70 use cases in a PowerPoint and then roll them out over 18 months. You start with 5, learn, adjust, and scale up progressively.
Where it shines: organizations already mature in agility, teams accustomed to sprints, test-and-learn culture. If your teams already do Scrum or SAFe, IAgile fits naturally into their existing rituals.
Where it struggles: I will readily admit it — if your organization has no agile culture, IAgile is hard to deploy. You are asking people to simultaneously adopt a new way of working AND AI. That is a double change, and it is sometimes too much. In that case, AUGMENT is probably a better starting point.
AUGMENT: acculturation above all
The AUGMENT framework, proposed by Thiga, follows 6 sequential steps: Awareness, Understanding, Generating ideas, Mastering tools, Embedding in processes, Nurturing the culture. The logic is clear: before deploying anything, make sure people understand what AI is, what it can do, and what it cannot do.
I underestimated this approach for a long time. I used to think acculturation would come naturally with usage. I was wrong. On an engagement with a 15,000-person industrial group, we launched ambitious AI POCs as early as month 2. Technically flawless. Adoption: 12%. Because the teams were not ready, did not understand why they were being asked to change, and were afraid for their jobs.
AUGMENT provides a framework for this foundational work. Each step has objectives, deliverables, indicators. It is methodical, structured, unifying.
Where it shines: large organizations with little digital culture, traditional sectors (manufacturing, construction, mutual insurance), contexts where fear of AI is strong. The "understand first, do later" approach significantly reduces change resistance.
Where it struggles: time. AUGMENT is a 12 to 18-month journey before arriving at tangible results. When your executive committee wants results by the quarter, explaining that you are still in the "Understanding" phase at month 4 is a perilous exercise. And frankly, some teams learn better by doing than by attending training sessions.
AI Ladder: data as the foundation
IBM's AI Ladder is historically the best-known framework. Four rungs: Collect (gather data), Organize (structure and govern), Analyze (model and train), Infuse (integrate into business processes). The ladder metaphor is compelling: no shortcuts, you climb rung by rung.
The approach has merit. Data quality is indeed the topic everyone underestimates. I have seen companies invest 2 million in a magnificent predictive model, fed by inconsistent, incomplete, duplicated data. Result: a model that is accurate on false data. Garbage in, garbage out — but with neural networks.
Where it shines: data-heavy organizations (banking, telecom, retail with large transaction volumes) that have a genuine data quality problem. For these companies, spending 6 months structuring their data lake before touching machine learning is not wasted time — it is a prerequisite.
Where it struggles: the world has changed. The AI Ladder was designed in the era of classical machine learning, where data was the primary bottleneck. With LLMs, a company can deploy useful generative AI without having solved its data quality problem. So requiring all 4 rungs in order is sometimes excessive. And let us be honest: it is also a framework that naturally leads to the IBM ecosystem — Watson, Cloud Pak for Data. The commercial bias is transparent, but real.
Side-by-side comparison
| Criterion | IAgile | AUGMENT | AI Ladder |
|---|---|---|---|
| Philosophy | Iterative, agile-native | Sequential acculturation | Data-centric, linear |
| Typical duration | 3-6 months for first results | 12-18 months full cycle | 6-12 months before "Infuse" |
| Organization prerequisite | Existing agile maturity | Leadership sponsorship + training budget | Data infrastructure, data eng. team |
| Primary focus | Concrete use cases, fast feedback | People, understanding, adoption | Data quality, ML pipeline |
| Main risk | Moving too fast for the organization | Too slow, loss of momentum | Rigid, IBM technology bias |
| Ideal for | Scale-ups, consulting firms, agile orgs | Large traditional enterprises | Data-heavy industries |
| Governance | Integrated into agile rituals | Dedicated acculturation committee | CDO / data office |
So which one should you pick?
My answer, after having used all three in different contexts: start from your organization, not from the framework.
If you are an agile organization (sprints, retros, feedback culture), IAgile fits naturally. It is the approach I detail in the dedicated article, and it is the one I use at inspearit. But I do not recommend it if your organization is still discovering Scrum.
If your teams are afraid of AI, if change resistance is strong, if the words "artificial intelligence" trigger defensive reactions — start with AUGMENT. Acculturation is not a luxury, it is a prerequisite. I learned that lesson the hard way.
If your fundamental problem is data quality, if your data teams spend 80% of their time cleaning rather than modeling, the AI Ladder lays the right foundations. Use it without necessarily buying the IBM stack that comes with it.
And in real life? You mix. On my last long engagement, an 8,000-person insurer, we used AUGMENT logic for the first 3 months (acculturation, manager training with the M3K framework), then switched to IAgile for execution. The glue between the two was clear AI governance with a committee that decided fast.
No framework will save you if your governance is vague, if your sponsor is not engaged, or if nobody wants to stop the projects that are not working. The framework provides rhythm and structure. The rest is leadership.