Building an AI-Forward Consultancy

An Elusive Goal

How can a consulting firm successfully adopt AI for itself? And then, how can it productize that capability in a way that is not just competitive, but a profound differentiator in the market?

Let's start by outlining the problem space here. Why is this even a necessary question? The most obvious driver, dwarfing all others, is simply that everyone else is doing it. There's a universal push, an undeniable gold rush, a massive wave building on the horizon. Similar to digital transformation, Agile, or the move to the cloud, a company feels it has to participate, regardless of how skilled it is or how much tangible value is truly there. The only alternative is to take a contrarian stance, to specifically call out the lack of value or the potential pitfalls.

Think about the Agile movement. A consultancy can take one of two positions: "Here's our unique, differentiated approach to Agile, and we're the best at it, so let us do it for you," or, "Here's why mainstream Agile is a mess, and you should listen to our superior alternative." Either way, you are forced to engage with the concept to differentiate yourself. You have to swim in that current.

A New Kind of Problem

However, a key aspect of this new problem space makes it fundamentally different. With Agile, cloud, or digital transformation, the path is relatively known. You can hire certified individuals, build a practice by bringing in good people, and partner with established organizations like the Scaled Agile Alliance. You can build a network for lead generation, assemble a delivery force, and make it happen through sheer force of will. If you throw enough people and money at the problem, you will get a result.

AI is not analogous to that today. It is an emerging technology, not a settled one. There are very few established standards or long-term career paths that guarantee expertise. Let me flesh this out. Let's say you hire someone with a Ph.D. in AI and machine learning who has been working in the space for 25 years. That, by no means, guarantees or even suggests that they will possess relevant thought leadership for enterprise AI architecture or adoption in 2025. Until two or three years ago, this technology was more of a science, a research and development discipline. The ground has shifted so completely that even someone who graduated last year with a computer science degree focused on AI cannot lean solely on what they know.

This is perhaps a unique facet of AI compared to other technologies: it is progressing at an insane speed. Major providers are constantly leapfrogging each other, and hundreds of billions of dollars are being injected into the market. A recent graduate’s knowledge, if not continuously updated, will be outdated. It is not a settled market. This leads to a crucial conclusion: you must have a think tank, a continuous motion of research, development, and experimentation, at the very core of your offerings and your internal adoption strategy.

The ROI Dilemma

The market fundamentally misunderstands AI, treating it as a tool (like Jira) or a solution (like Agile). The inevitable conclusion, however, is that AI is a capability, much like "agility" itself, an inherent capacity for intelligent automation. This distinction is critical. Clients demand ROI from scalable "AI solutions," but this clashes with the consulting model because the underlying technology is a chaotic, moving target. Without a constant R&D motion to cultivate the capability, a firm’s expertise evaporates as quickly as their committed-to techniques become obsolete.

This reality mandates a new principle: a commitment to cultivating capability. A long-term AI strategy cannot be premeditated; it must emerge from a deep, hands-on understanding of what is possible. This requires building modular, vendor-agnostic architectures that can adapt, rather than locking into any single technology.

Firms that miss this distinction are trapped in a paradox. They must project authority in the market, but their business model, built on large-scale, high-ROI engagements, is misaligned with a nascent capability. This leads to brutal sales cycles and a culture clash between the promise of innovation and the reality of quarterly targets. "AI as a Solution" is not a viable service line in its current state. If you doubt this, look no further than the legion of “AI Wrapper” startups, companies that built thin applications on top of foundational models. Startups like Jasper, once valued at over a billion dollars, saw their core value proposition evaporate as the underlying models became more powerful and directly accessible. We're witnessing an ever-more-crowded graveyard of AI-powered copywriting tools and summarizers that attracted massive valuations before crashing out, unable to build a defensible moat against the very platforms they relied on. This is because they weren't selling a capability; they were selling an opinionated solution in an emergent problem space.

An Emergent Problem Space

The fundamental challenge, then, is recognizing AI for what it is: an emergent problem space defined by unprecedented velocity. For consultancies, this speed isn't a threat, but a massive opportunity. It means that the key to unlocking value and achieving real problem-product fit lies in a deep, continuous understanding of the market. The firms that succeed will be those that can adapt to this pace, not by chasing every new development, but by staying grounded in the evolving needs of their clients and identifying where new capabilities can deliver tangible results. Success in this space is less about having a static solution and more about having the market awareness to apply the right capability at the right time.

This presents the central strategic question: How can a consultancy fund the continuous R&D necessary to lead the market, and how can it structure that investment to generate recognizable value rather than just becoming a cost center with cool demos?

Drink Your Own Kool-Aid

The answer is you do it all internally. You commit to drinking your own Kool-Aid as the first order of business. You build an AI Center of Excellence, and you set the immediate expectation for return on investment to come from internal gains. The initial value of hiring AI experts and giving them a budget should be measured in internal cost savings and revenue boosts from adopting AI tools yourself. Perhaps over a one-to-two-year pipeline, you expect 75% of the net gains to come from this internal adoption. Then, after that two-year mark, the balance can flip, with 75% of the gains shifting to external revenue generation, go-to-market strategies, and product sales.

Larger firms, facing immense pressure to have an AI offering, often attempt a dangerous shortcut: selling the strategic vision without undertaking the arduous internal transformation. And, to be frank, that is understandable! Transforming a massive, culturally dense organization for any change is a multi-year slog, let alone for a volatile emerging technology like AI. But in trying to cut this corner, they cut themselves off at the knees. They end up selling a capability they haven't cultivated, a product they can't scale, and a solution they can't truly implement. This gap between their sales pitch and their delivery capability becomes a chasm, leading to failed projects and a tarnished reputation. This self-inflicted wound creates a massive opportunity for smaller, more agile firms. For them, a genuine, lived-in commitment to AI-augmented work isn't just a talking point; it's a profound competitive advantage, allowing them to deliver on promises their larger competitors can only make.

A Culture of Experimentation

Fortunately, the principles required to navigate this uncertainty are neither new nor mysterious. They are the well-worn fundamentals of any successful transformation. The vision for an AI-first future must be backed by a real commitment. This means a financial commitment, yes, but more importantly, a cultural commitment to disciplined experimentation. This means calmly applying the proven 'build, test, learn' cycle, granting teams the freedom to fail, and treating every misstep not as a disaster, but as a successful experiment that yields critical learning.

In any emerging technological space, the core commitment cannot be to specific outcomes, but to the approach: a relentless process of learning and acceleration, which is standard operating procedure for innovation. When this disciplined commitment exists, the path forward becomes clear: an internal adoption team captures learnings from both successes and failures, transforming that direct experience into the raw material for authentic, proven products. This team then leads the charge not with theory, but with lived-in authority.

A pragmatic, agile way to begin this journey is with fractional leadership. This isn't about boiling the ocean, but about taking a sensible first step. Bringing in expertise for a defined period allows a firm to de-risk the initial investment, conduct a proper readiness assessment, and begin the familiar work of discovery, all without committing to a full-scale transformation from day one.

The Path of True Transformation

When I talk about transformation, I'm not talking about simply buying a new piece of software. I'm talking about a holistic shift across four crucial areas: Training, Ways of Working, Technology, and finally, Product/GTM.

1. Training

Everything must start with capabilities. There is so much training out there that just regurgitates what AI is and what people say it can do. This is insufficient. People need to get hands-on. They need to fail. They need to learn how to work with the technology themselves, on tasks that are not part of their primary daily responsibilities. This requires a commitment of time. I regularly develop custom training content, including full-fledged labs with a wide array of AI tools, for exactly this purpose.

The goal is to give people hands-on experience and, while they're gaining it, give them the language to understand what they're experiencing. By way of example, it’s crucial to clarify the concept of hallucination. A transformative insight for many is that, in a sense, all a large language model does is hallucinate. That is its function; it is a hallucination machine. The only difference between what we call a "hallucination" and a "correct answer" is that you, the user, apply your ontological understanding of the world and judge the output as true or not true. From the LLM's perspective, the process is technologically identical.

Understanding these nuances, backed by experience, gives everyone in the organization the authority to speak into the market. Much of the hesitancy and weirdness around AI today comes from the fact that you can tell most people talking about it have no real authority; they're just echoing a research paper someone else wrote. Giving your people hands-on experience is the antidote. The core motivator for this investment is the acknowledgment that AI is here to stay. We may not know which specific platforms will win the global arms race, but we know it's not going away. Just like learning to use a computer or the internet, people need to learn to use this. The best professional investment you can make is to gain serious, hands-on experience, not just certifications, with generative AI.

2. Ways of Working

The second pillar is a re-evaluation of how we work. So many of our business frameworks, methodologies, and processes are designed to manage constraints that AI is actively removing. If you adopt AI, you gain not just the ability, but the imperative to work in different ways.

Here’s a thought experiment I like to use:

For a moment, bracket out everything you think you know about large language models.

Imagine you have a machine capable of simulating human reasoning with 80% accuracy. That's all it does. How would you deploy and integrate that machine into your business to achieve your goals?

Now, what if it was only 60% accurate? How would that change your approach?

What if it was 80% accurate for some use cases but only 40% for others? What if it was 80% accurate most of the time, but for the 20% of tasks that are most critical (the most damaging if wrong and the most rewarding if right) its accuracy was much lower?

Asking these questions gets people thinking about how to deploy what AI actually is: a reasoning and speaking simulation machine. We can do incredibly creative things with it, but it takes deliberate thought. It's helpful to discard the conventions we've heard, because many of them are misinformed, oversimplified, or magical thinking.

Consider a development team working in Sprints that roll up into an Agile Release Train, all culminating in PI Planning. What happens to that entire structure if the teams adopt AI-augmented development and their velocity triples? The amount of information they need to account for in planning meetings explodes. The old ceremonies break under the pressure of unprecedented throughput. We need to change our ways of working to manage the constraints that actually exist, like the increased technical debt from AI-generated boilerplate code, and not the ones that have been rendered obsolete, like an arbitrary number of story points a team can complete in a Sprint.

3. Technology and Tech Stack

Only after addressing training and ways of working should you focus on the tech stack. People often want to start here, but implementing tools without the foundational cultural and process changes is a recipe for disaster.

When we do get to technology, we're talking about creating a common-sense policy for AI tools, establishing a list of approved platforms, and forming strategic partnerships with key vendors. This involves making critical architectural decisions. For instance, do you go directly to OpenAI or Anthropic for API access, or do you leverage your existing MSA with Microsoft and use Azure OpenAI Service for GRC compliance? Or perhaps you use AWS Bedrock's version of the Claude model. These choices have significant implications for governance, risk, and compliance, especially when dealing with PII, HIPAA, or ITAR requirements. The major infrastructure providers are becoming the picks-and-shovels suppliers for the AI gold rush, and choosing the right one is a crucial part of adoption.

From there, you move to the integration layer, exploring techniques like Model Context Protocol, agentic AI, and advanced enterprise search. You have to build a toolkit of ideas and technologies to not only adopt AI yourself but to eventually implement it for others. I would approach this unit by unit. For developers, how can AI conduct code reviews, integrate into their IDEs, and orchestrate a team of AI coding agents? For salespeople, how can they leverage models to analyze sales calls from Gong, prioritize opportunities in their CRM, or automate outreach? For IT, how can AI help resolve incidents faster? For HR, how can it improve documentation and onboarding?

I use a simple model for this: Reflect, Inflect, Inject.

• Reflect: Think about what you're currently doing and who you're doing it for.

• Inflect: Modulate your ways of working and your tools to support a new, desired outcome.

• Inject: Meaningfully insert AI capabilities into that new process.

4. Product

After all of this, after the training, the new ways of working, and the technological integration, then you have a product. You finally have the opportunity to productize your experience because you have seen it in action. Everyone is starting from the same proven foundation. This marks a strategic shift from solution-based selling to capability-based selling. It is a more viable path for firms seeking to capture value from the AI market in 2026 and beyond, as you are not selling a fixed product, but a dynamic, proven capability.

“This marks a strategic shift from solution-based selling to capability-based selling. It is a more viable path for firms seeking to capture value from the AI market in 2026 and beyond, as you are not selling a fixed product, but a dynamic, proven capability.”

If you try to productize before this internal transformation, you're in a precarious position. Unlike Agile consulting, where you can always hire another certified consultant to fulfill a statement of work, that safety net doesn't exist in AI. There's no standard set of expectations, and past job experience is not a reliable indicator of current competence. You can't just write up a magical SOW, promise a client their wildest desires, and then staff it from the open market. The field is saturated with snake oil.

The alternative, to say, "We're not touching it because we're not sure how," is equally non-viable. The market wants AI, and it will go to the people who can provide it. The critical question of our time is: can you deliver on the promise? And the circle of those who truly can is much smaller than we might have originally thought.

A Case Study: AI-Augmented Migrations

Let's ground this with a real-world example. Imagine you're an Atlassian consultancy. One of your major revenue streams is Atlassian Cloud Migration. Atlassian has announced the end of life for their Data Center products, creating an urgent, closed market with a finite amount of money and time. The pressure is on to capture those migration deals. Your migration team is likely maxed out. You can hire more technicians to write Python scripts, coax the Jira Cloud Migration Assistant, analyze plug-in compatibility, and conduct UAT, but it's a tremendous amount of manual, high-cognition work.

The hard constraint is your team's bandwidth. Now, what if you go in and transform the way that team works? You're not selling an "AI product" to the client; you're selling an Atlassian Cloud Migration. But internally, you're training your team to write those Python scripts 30% faster with AI assistance. You're automating customer communication plans using an AI Agent with access to a Gmail inbox and the project’s Jira board. You're using AI to conduct detailed plug-in compatibility research and to automatically convert large CSV files from one configuration standard to another.

The client doesn't get an "AI migration." They just get their migration, only it's completed faster. Your company gets it done with a lower cost of goods sold, which means more margin. And you can take on more migrations because you can execute them faster and more reliably. That is AI adoption with tangible ROI for everyone. The only thing you're giving up is the shiny "AI" label on the SOW. You're not selling an AI product; you are the AI product. Your consultancy becomes AI-forward, empowered by new ways of working.

This internal transformation does more than just save money; it builds an operational moat. Your whole company starts to develop a deep market awareness. Your ability to recognize opportunities, prototype products, and move at a speed that was previously unimaginable becomes a core competency. When people see how these tools can make their jobs easier, they become motivated to experiment and stay on the cutting edge. It democratizes innovation.

Once your company operates this way, anyone can spin off products based on their expertise. You can sell AI Agile Coaches as a Service. You can offer "95% Automated Atlassian Migrations" as a Marketplace app. You can deploy agentic AI JIRA Administrators to handle service requests. You can build robust, future-proof Enterprise Knowledge Graphs. You can form meaningful partnerships with the true leaders like OpenAI and Anthropic, not just as a customer, but as a genuine implementation partner and thought leader. And perhaps most importantly, you stand poised to engage existential threats to the Atlassian ecosystem because your expertise is more widely located in “Enterprise AI Architecture” than in simply knowing how to turn on Atlassian Rovo. Internal transformation opens the door to market dominance.

The AI-Forward Philosophy

The path forward, then, is not about choosing the right tools, but about adopting the right philosophy. It begins with the conviction that AI is a capability to be cultivated, not a product to be sold. This requires a disciplined, internal-first approach, where the initial ROI is measured in your own operational efficiency and enhanced ways of working. By drinking your own Kool-Aid, you build the authentic, hard-won expertise that the market is desperately seeking.

This is how you create an operational moat, not with a specific technology, but with a culture of continuous learning and a team that has genuinely experienced a better way of working.

It is a commitment to substance over sizzle, and it is the only sustainable path to market leadership.