In the midst of today's atmosphere when everyone is gung-ho about AI, few if any are speaking of the specifics of AI adoption and implementation inside companies. AI is not a magic wand—it needs to be fit into the organisation, its processes, and made a part of the life of the people who populate the organisation. It won't get its due just by hyperbole and constant focus on the promises of AI. Someone, and a lot of someones, has to get down to the hardcore practicalities of using it. Individual AI usage ≠ organisational AI implementation. A company can buy an excellent AI tool and still fail because it doesn't know how to integrate it into actual workflows.
People don't resist AI merely because they fear AI. They resist disruption to functioning processes. And therefore: the first implementation objective should not be transformation. It should be integration. That creates a two-stage implementation model:
Stage 1: Fit AI into the existing process.
Stage 2: Redesign the process once adoption is established.
That is the problem I see in companies nowadays, both customers and vendors. People are simply not willing to put their nose to the nitty-gritty of applying AI to internal realities. Even as a customer when I have spoken to vendors, I have found it close to impossible to get them to apply the AI to my use case, with the answers being high-point and highlight-based rather than specific and actionable. That is, in my opinion, both a mistake as well as something that is to be expected given that this field is so new.
The Need Of The Hour
Now the need of the hour is for vendors to get practical, and start actually speaking in terms of specific implementation points inside workflows and processes—and quit their grandiose statements on their websites, brochures, customer communications, and in their discovery calls. That means getting a deep understanding of how, what, where, when, and why their tools are being or can be implemented inside companies' existing systems.
Shifting the Lens: Practical Application Across Marketing
I have consistently taken the position that AI is useful. It is just how we are using AI that is a problem, alongside certain attendant risks. Shifting the lens diametrically opposite, we must look at how AI is actually being used and what advantages we are getting. Taking marketing as a function, there are three distinct use cases where I have been using AI:
1. content for writing, copy and scripting
2. strategic research, and
3. Go-To-Market (GTM) planning & execution.
To take value out of AI, one must follow specific methods to ensure performance, such that the result fits the output that is desired – and should be suited to the task type at hand. In content, this spans translation and writing technical material. In strategic research, it involves deep context setting. In GTM, it includes tools like AI dialers, WhatsApp, and lead handling.
My Credentials
People often just copy, paste, print, or publish. That is exactly how you should not be using AI; that is where the risks emanate from. My credentials in content come from translating entire books from one language to another, creating almost 3,000 slides based on my work, and preparing social media posts for seven to eight full commentaries of the Bhagavad Gita over the past eight to nine years, using tools ranging from Google Translate and online dictionaries like Shabdkosh to Gemini, Claude, Grok, and ChatGPT. In Strategy, I have been involved in preparing strategies for organisations for over 7 – 8 years now, and in GTM, my past includes a regional hub head position as head of sales.
The Basics Of How
There are two aspects of the how – actually using the tools – which is a topic unto itself, and how the tools themselves can be successfully implemented and adopted inside organisations. The first is the easier part of the story – the second, unfortunately, is a story unto itself requiring deep attention. Before I go deeper into the actually using part, let me just quickly introduce the tougher part- getting people to use it is not simple a matter of objection handling, or fear, or lack of clarity, etc.
It is a question of process – this is where all vendors get it wrong; the problem is existing processes are already operational inside companies – you cannot get people to scrap them. That is a recipe for disaster. The only way is to ensure your AI tool or solution fits existing processes with as little disruption as possible, and then slowly, over time, as usage and comfort evolves – the customers will themselves evolve their processes to capture true power of the tools once they are using it.
The only caveat – there will be failures in those organisations that demand instant or quick ROI. That is why judgement of the customer company is so crucial, so you can build in quick wins in the resulting implementations. Quick wins are still necessary—not because they prove the ultimate value of AI, but because they create organisational confidence and buy-in.
The Specifics Of The Use Cases, And The Process Of Use
Content and Translation: Brain-Level Verification Over Copy-Paste
I first give my premise to the AI in detail and then get into a long conversation to arrive at the desired output. For translation, I do not just translate. I compare line by line and word by word, going into the semantic and syntactical meanings of words, verifying against physical copies of the Bhagavad Gita to make sure the content is understandable and accurate. There is a lot of work required at the brain level and physical activity level. Furthermore, for research-driven content, I lay out authentic sources, limit the AI to specific inputs like newspaper articles, and ask it to extract information with verifiable links.
The challenge inside companies in this process is the realisation that it is not cut and paste, and that domain knowledge of the task at hand is mandatory. This has to happen at the initial introduction stage, when the initial user team is finalised. Vendors have little say in this – but a simple advisory around this, asking them to have domain experts who are also avidly interested in new tools and technology which would be a definite plus towards ROI realisation.
Strategic Research: Grounding AI with Real-World Intelligence
Strategic research is a deeper topic that requires significant detailing. The initial steps are the same as in a non-AI environment: the problem definition, customer brief, background understanding, core research using website, site / store / field visits, customer calls etc does not change- this is vital. Nothing in the AI world can be a substitute for this. Absolutely nothing. Further, AI is excellent at expanding and interrogating a knowledge base; it is much less reliable at determining whether the knowledge base itself reflects reality.
I establish the context of the strategy by providing the industrial background, giving full customer or task brief details, giving the AI specific directions, and detailing the exact points required out of it. After the first strategy is generated, I go deeper by conducting my own research using Google search or physical visits in the market to provide deeper context. When AI comes out with an inaccurate competition list, I manually override it, provide the actual real-world competition, and instruct the AI to research based on those true parameters. A lot of work goes into implementing proper strategic research functionality within marketing. For more on this – follow me, when I write a dedicated article on how to strategise using a blend of AI and traditional approaches.
Go-To-Market (GTM): Implementing Supervised AI
The toughest use case- this one actually involves teams. And here, buy-in is essential. Further, as this is the first use case we are discussing that is dealing with multi-step, multifunction, and maybe even multi-organisational (in B2B companies) touch-points – it also involves complex processes. You need buy-in from all the touchpoints involves – going topdown is useless here. You need to understand their flow of the lead from initial contact all the way to closure and to customer service functions, what data is recorded, what conversations are conducted in what language and by whom. Again – please follow me, for getting the article that goes deep into this process.
In GTM, first of all you need to choose where you are going to start from – this is the revenue engine of the company, and needs to be carefully planned, not just jumped into. AI Vendors should be particular on this point – it is for you to educate the customer on this invisible point. In that sense, there are multiple starting points within a GTM for AI to start - it can be applied as a voice dialer, in inward lead handling, outbound calling, or lead handling through WhatsApp to name but a few. I have tried all. And where to start is fundamentally dependent on what the customer wants to achieve, yes – but also what are the challenges facing that particular point or function? The actual implementation will fundamentally depend on this answer – not one vendor I have spoken goes this deep.
Going ahead, implementing this effectively requires repeated training and internal testing before launching. Once conversations start happening in the initial phase, you have to sit and monitor each conversation. This is Supervised AI Deployment. If it is a WhatsApp conversation, you need to read every single line of every conversation, correct the AI where it strays, set new rules, and establish guard rails. It has to be properly implemented, and the challenge today is that people simply do not know how to execute that. This Supervised AI Deployment is the period in which humans actively monitor live AI interactions, identify failures, correct the system and progressively establish rules and guardrails.
Conclusion
In conclusion, AI is less about the tool itself and far more about the processes it has to blend into, the people who will use it, and the organisation in which it has to operate. AI does not exist in isolation; it has to become part of the organisation’s living processes if it is to deliver sustained value. And that is where AI vendors usually go wrong. After implementing AI in a few different environments, my biggest learning has been simple: the AI or technology tool has to blend seamlessly into the way people actually work. If it does not, it will not get used. I have seen this fail abjectly in practice.
So, for now, my message to AI vendors is straightforward: spend considerably more time understanding the customer. Understand how they work, how they intend to use the AI, the contextual setting in which it will be used, the processes it has to fit into, the alternatives available to them, and the people who will ultimately have to use it. Unless you understand the customer in depth, there will inevitably be problems in implementation, adoption and, ultimately, scale. The AI adoption gap is therefore not necessarily a technology gap. It is an implementation gap. And that is the black hole that needs to be addressed.
Comments
Post a Comment