We’re in the middle of a transformation—but are we really transforming?
Friday hot take: When it comes to AI and data, many organizations are stuck in a “faster horse” mindset, optimising legacy workflows rather than rethinking what’s possible.
This reflection was sparked by a post from Bethany Lyons about cloud migrations and how too many companies simply port old logic into new platforms without fully embracing the capabilities of the cloud. It struck a chord—because the same thing is happening in AI adoption.
The Legacy Data Workflow (aka the Faster Horse)
Take this common scenario:
A sales team needs a new forecast metric.
Traditionally, this kicks off a chain reaction:
The data team sources data from multiple sales systems
It passes through warehouse layers, transformation models, and validation rules
Eventually, the business analyst builds the metric into a dashboard
Best case? You’re looking at weeks or months.
Some modern tools (like Paradime) streamline parts of this pipeline—but the architecture remains the same. You’re still bolting AI onto a traditional, human-driven process.
Rethinking the Model: From Pipelines to Prototypes
What if we stopped trying to optimise the old flow, and instead reimagined it entirely?
Enter AI Digital Twins powered by Data Object Graphs (DOGs).
Let’s rethink that sales metric:
AI agents interrogate the sales systems directly
They identify only the relevant data
Then automatically generate the pipeline, validation rules, logic, and even a visualisation (e.g., a donut chart)
With light human review, the result is deployed across apps, APIs, and dashboards
All of this happens in minutes, not months.
No wrangling with brittle ETL chains. No waiting on backlogs. Just delivering insight at the speed of thought.
Beyond Automation: It’s About Redefining the Outcome
This isn’t just about AI doing what humans used to do, but faster.
It’s about asking: Do we even need to do it that way anymore?
Rather than optimising for traditional deliverables, we can focus on end-to-end business outcomes, using AI to:
Automate metric creation
Build user interfaces dynamically
Simulate business processes before rollout
Drive real-time integration with transactional systems
And yes—some of the work still falls into the 20% that's complex, messy, and human. But the 80% that isn’t? We’re burning cycles on redundant variations of the same tasks.
Agents on Recon Missions
This approach extends even further when we embrace agent-based models.
Imagine a network of agents that:
Explore your systems like reconnaissance scouts
Understand your processes even without full documentation
Negotiate with other system agents (think: ecosystem-level orchestration)
This is where AI twins, DOGs, and frameworks like MCP come into play—giving us ecosystems of intelligence, not just smarter tools.
Crossing the Chasm (Again)
AI isn't just a faster engine. It’s a new mode of transport.
But to adopt it effectively, we must leave behind the comfort of known workflows—even if they’re just a year or two old. It takes courage, vision, and often a controlled safe space to prototype new workflows before transforming at scale.
As with all big shifts, there's a psychological gap. We’re not just automating the past—we’re designing the future.
With Dataception Ltd’s DOGs and AI Digital Twins, the journey from data to decision is no longer a pipeline—it’s a conversation.
Let’s stop feeding oats to a faster horse and start building the car. 🚗