Suggested category: AI / Automation
Suggested author: Arthur Marinis
Suggested read time: 5 Minute Read
Source: Webinar with DingGo, "Beyond the Hype: How Strategic AI Drives Real ROI"
AI adoption has moved quickly from curiosity to board-level priority. The problem is that many organisations are still treating AI as a technology project first, rather than a business improvement discipline.
That distinction matters.
During our recent webinar with DingGo, we explored what happens when AI is tied to clear business outcomes instead of vague experimentation. DingGo's journey is a useful example because it is not about chasing every new model release or rebranding the company as an AI business. It is about using AI to improve the way an existing business operates, serves customers, manages cost, and builds a stronger product.
The first question should not be "How do we use AI?"
It should be: what business outcome are we trying to improve?
For DingGo, one of the clearest goals was reducing cost to serve. That gave the team a practical filter for decision making. Instead of spreading effort across disconnected experiments, they could assess each AI initiative against a real operational target.
As discussed in the webinar, AI needs to be aligned to business value. It is not enough to say "we want to use AI". The use case has to connect to a measurable result: lower cost, faster service, better customer experience, improved accuracy, new revenue, or reduced operational friction.
Without that alignment, AI quickly becomes a hammer looking for a nail.
Many companies underestimate the role of data quality, ownership, and governance in AI success.
AI systems do not become useful simply because data has been uploaded to them. The data needs to be relevant, structured, governed, secure, and appropriate for the use case. It also needs to be handled in a way that does not expose the business to privacy, compliance, or operational risk.
This is where guardrails become important.
Guardrails are not just a security feature. They shape what goes into an AI system, what comes out of it, how responses are evaluated, and where human review is required. For production AI, especially in regulated or customer-facing environments, security, accuracy, governance, and auditability need to be designed in from the beginning.
One of the most interesting parts of DingGo's journey was the decision to pivot.
The team initially looked at building its own AI model from its unique vehicle damage, repair, and cost data. That made sense at the time. DingGo had domain-specific data and a clear use case.
But the model landscape moved quickly. Large language models and computer vision models were improving faster than a single business could reasonably maintain on its own.
Rather than continuing down a custom model path for the sake of it, DingGo shifted towards using its people and proprietary data as the differentiator. That meant human-in-the-loop workflows, model correction, fine-tuning, and feedback loops.
This is a practical pattern for many businesses. The competitive advantage is not always the model itself. Often, it is the workflow around the model: the quality of the data, the way humans validate outputs, the domain expertise embedded in the process, and the way the system improves over time.
DingGo made an important point in the webinar: they are not trying to become an AI company.
That is the right mindset for many organisations.
AI is not the product in every business. Sometimes it is the capability that improves the product. It can elevate the customer experience, increase operational efficiency, and unlock new features, but it still needs to serve the business model.
For SaaS and product companies, this is especially important. Investors, boards, and customers may be excited by AI, but the value comes from applying it to differentiated business capabilities. In DingGo's case, that means combining unique data, human expertise, and automated workflows to improve outcomes in a complex repair ecosystem.
The AI market is noisy. New tools, models, frameworks, and agentic patterns appear constantly. It is easy for teams to feel they are either moving too slowly or missing something important.
That fear can lead to rushed decisions.
Businesses need a way to separate exploration from execution. It is healthy to stay close to new AI capabilities, test new models, and understand where the technology is heading. But production systems need a different level of discipline: evaluation, security, cost control, rollback plans, monitoring, and clear ownership.
The goal is not to adopt every new model. The goal is to create an architecture and operating model that allows the business to change models safely when there is a reason to do so.
AI adoption is not only a technical challenge. It is a people challenge.
When employees hear that AI is coming into their workflow, many assume the objective is replacement. That reaction is understandable, especially when the broader public conversation around AI is often framed around job loss.
The better message is more specific and more human: we want to find the most frustrating, repetitive, low-value parts of your job and reduce them so you can spend more time on work that uses your judgement and skill.
That does not happen with one announcement. It requires repetition, education, involvement, and visible examples. Teams need to see AI as something they can shape, not something being done to them.
Internal hack days, guided experiments, and stakeholder mapping can help. They give people a safe way to experience the technology and connect it to their actual work.
The strongest AI programs tend to start small.
That does not mean thinking small. It means choosing a focused use case, defining the metrics, putting the right guardrails in place, testing the workflow, and learning before scaling.
Once the first use cases prove value, the next challenge is coherence. DingGo described having more than 30 AI initiatives underway. That creates a new need: a strategy that connects prompt management, measurement, testing, model routing, retrieval, security monitoring, and governance.
This is where AI starts to look less like a set of experiments and more like a managed business capability.
DingGo's experience highlights several lessons for any organisation moving from AI experimentation to AI value:
AI can create real ROI, but only when it is connected to the way the business actually works.
The organisations that win will not be the ones that use AI everywhere. They will be the ones that apply it deliberately, govern it properly, and use it to strengthen the capabilities that already make them different.
base2Services helps organisations design, build, and operate AI solutions with the right foundations in place: secure cloud infrastructure, data readiness, guardrails, MLOps, governance, and production support.
If you are exploring how AI can reduce cost, improve operations, or create new product capability, our AI Factory approach can help you move from idea to production with confidence.
Get in touch with base2Services to start shaping an AI strategy that is practical, secure, and aligned to real business value.