What are DOGs?

Redefining Analytics with Data Object Graphs

Data Object Graphs (DOGs) unify data and processes into a single, scalable framework, empowering organizations to turn raw data into actionable insights. By combining advanced analytics with intelligent automation, DOGs streamline workflows and enable rapid, reliable decision-making. Discover how this innovative approach bridges traditional analytics with the evolving demands of AI, delivering efficiency and scalability at every stage of the data lifecycle.

In today's rapidly evolving technological landscape, traditional data processing models are being challenged by the complexities of modern business processes and the demands of artificial intelligence (AI) applications. Directed Acyclic Graphs (DAGs) have long been the backbone of data pipelines and orchestration tools. However, as AI systems increasingly rely on feedback loops and complex data structures, there's a growing need for more sophisticated models.

This article introduces the concept of Data Object Graphs (DOGs), a hybrid data and execution graph model designed to address the limitations of DAGs. Developed by Dataception Ltd, DOGs blend data and execution nodes to create queryable, adaptable, and traversable graphs suitable for both AI and traditional analytics use cases. By integrating methods and state within nodes, DOGs offer a dynamic and executable framework that mirrors intricate business processes, supports AI agents, and handles complex data types.

We will explore how DOGs are revolutionizing data and AI development across various parts:

  1. Moving Beyond DAGs : Understanding the need for DOGs in complex workflows
  2. Introducing Agent DOG : How AI Agents Interact with Data Object Graphs
  3. Multi-Agent Data Object Graphs : Collaborative AI agents working together
  4. PackRunner Architecture : The infrastructure supporting DOGs and AI agents
  5. Digital Twins of Business Processes : Simulating processes with DOGs
  6. Enhancing Decision Intelligence : Combining Agentic AI with DOGs
  7. Executable Metric Trees as DOGs : Practical applications in BI and metrics
  8. Pure Agentic vs. Model-Accelerated Workflow : Balancing automation and human collaboration
  9. Walking the DOG : Agentic Data Object Graphs—A Query Plan for Your Business
  10. Sniffing out the Map : Transparency in AI Decision-Making — DOGs Agentic "Query Plans Process"
  11. Follow the Path or Chase the Squirrels?  Agentic Deterministic vs. Probabilistic Planning
  12. New Era of Rapid Innovation : AI Is the Printing Press Moment for Data and Technology
  13. Unleashing AI Value : From Ruff Terrain to Business Treasure
  14. Your Business Processes Have Gone to the DOGs : (And That's a Good Thing!)
  15. Roll Over, LLMs!  The Rise of Reasoning Language Models (RLMs)
  16. Release the Greyhounds : From Prompt to Business Process
  17. Whoooaaa!  Proof That the Operational-Analytical Divide Is Disappearing 🚀
  18. AI Breeds Better DOG : AI-Generated Data Object Graphs for Business Processes

By delving into these topics, we aim to showcase how DOGs provide a robust foundation for modern data processing needs, enabling organizations to innovate rapidly while minimizing risks and costs. Whether you're a data scientist, AI practitioner, or business leader, understanding DOGs could be the key to unlocking new potentials in your data and AI initiatives.

15. ​Roll Over, LLMs!

The Rise of Reasoning Language Models (RLMs)

As AI continues to evolve, a new frontier is emerging—Reasoning Language Models (RLMs)—a next-generation AI approach that could redefine how machines think, solve problems, and generate insights. Unlike traditional Large Language Models (LLMs), which rely on pattern recognition and probabilistic text generation, RLMs introduce explicit, verifiable reasoning paths, moving beyond single-pass black-box processing.

Why RLMs Matter

While LLMs are powerful for general text generation, their lack of structured reasoning makes them unreliable for complex, multi-step problems. RLMs, on the other hand, actively explore multiple solution paths, evaluate intermediate steps, and backtrack when needed—mirroring human problem-solving.

By leveraging structured exploration using trees, chains, and graphs, combined with value and reward models, RLMs can systematically verify progress, adjust when necessary, and produce more transparent and reliable outputs. This makes them ideal for tasks such as:

  • Mathematical proofs
  • Scientific analysis
  • Complex business planning
  • Long-term strategy development

The Trade-Off: Capability vs. Computational Cost

RLMs offer superior reasoning, but this comes with higher computational costs and greater architectural complexity. The decision to use RLMs vs. traditional LLMs is ultimately a trade-off between advanced reasoning capabilities and resource efficiency.

The Core Components of RLMs

1. Reasoning Scheme – The blueprint defining how AI structures its thought process:

  • Reasoning Step – The fundamental unit of reasoning.
  • Reasoning Structure – How reasoning steps are organized (Chains, Trees, Graphs, or Nested structures).
  • Reasoning Strategy – How the reasoning structure evolves over time.

2. Operators  – The AI’s toolset for manipulating and navigating reasoning paths:

  • Structure Operators – Modify the reasoning framework.
  • Traversal Operators – Navigate through the reasoning process.
  • Update Operators – Improve specific steps.
  • Evaluate Operators – Assess the quality of reasoning at each stage.

3. Models  – Systems that generate, evaluate, and reward different reasoning approaches:

  • Policy Model – Generates reasoning steps.
  • Value Model – Evaluates reasoning states.
  • Reward Model – Assesses step quality and guides optimization.

4. Pipelines  – The workflows that orchestrate the RLM’s operations:

  • Inference – Runs the trained model for real-world application.
  • Training – Continuously improves reasoning performance.
  • Data Generation – Creates structured training datasets.

The Future of RLMs and Data Object Graphs (DOGs)

There’s already a great paper, "Reasoning Language Models: A Blueprint" (link https://arxiv.org/abs/2501.11223), that lays out one approach to this concept. However, at Dataception, we’ve been exploring a similar but more efficient approach—using Data Object Graphs (DOGs) to tie all these reasoning components together.

Rather than relying on massive LLMs with 100Bs of parameters, we believe we can optimize reasoning processes within DOGs, making RLMs significantly more scalable and efficient. More on that soon… but one thing is clear:

AI isn’t just getting smarter—it’s learning how to think.

16. Release the Greyhounds

From Prompt to Business Process

Accelerating Business Transformation with AI Digital Twins

I had a fantastic session with Peter Everill discussing "Quantifying Your Value: The Framework Used to Realize £100M Profit" on Kyle Winterbottom’s Orbition Group podcast (link in comments). Peter outlined a consultancy-based framework for delivering end-to-end data products that drive profit directly to the P&L—sometimes to the tune of hundreds of millions.

The Key Takeaway?

It all starts with the business process.
(Yes! Every Data & Analytics team should be doing this!)

A key question that came up was:
"How can LLMs accelerate this process?"

At Dataception, we’ve built a solution that does exactly that—turning simple prompts into full end-to-end business processes using Data Object Graphs (DOGs).

How It Works: From Prompt to AI Digital Twin

  1. Capturing the Business Requirement
    • We use our Data Product Pyramid to extract the high-level requirement from stakeholders and convert it into a simple, structured prompt describing the ask.
  2. Generating the Business Process as a Data Object Graph (DOG)
    • Our AI-driven solution takes this prompt and generates a complete end-to-end business process, covering:
      • Data to Business Outcome
      • Defined Data Products (AI forecasts, LLMs—SLMs, OSS, and SaaS models—ML models, decision models, pipelines, metrics, and more)
    • The entire process is represented in a structured, human- and machine-readable markup language, where each data product is described alongside its required data.
  3. Automated Execution & Visualization
    • The DOG is executed using a graph and data product container-based distributed processing engine.
    • The entire process is visualized in an interactive 3D schematic, providing instant clarity into the workflow.

This is the AI Digital Twin in action—a fully executable, testable business process model.

Why This Matters: The Business Transformation Impact

  • Rapid Prototyping with AI & Digital Twins
    • We can test, iterate, and refine entire business processes in days—not months.
  • One-Click Execution at the Business Process Level
    • Individual data products can be run in isolation or executed as a complete end-to-end process.
  • Continuous Iteration & Adaptability
    • As business needs evolve, the DOG can be refined in real-time, incorporating data quality rules, validation, and automated testing.
  • Massive Speed & Low Risk
    • This approach allows businesses to "try out" AI and new ways of working in a risk-free environment before committing to full-scale deployment.

The Future of AI-Driven Business Change

This approach isn’t just faster—it’s safer, smarter, and more cost-effective.

By using AI Digital Twins, businesses can simulate, test, and optimize workflows in real time, unlocking true innovation without the traditional risks of large-scale digital transformation projects.

Want to see how it works? Get in touch with us if you’re interested in trying it out.

With Dataception’s DOGs (Data Object Graphs), AI really is just a walk in the park. 🚀

17. Whoooaaa!

Proof That the Operational-Analytical Divide Is Disappearing 🚀

The long-standing divide between operational and analytical systems is vanishing—if not already gone.

We just used our Data Product Pyramid Gen AI tool to create transactional Data Object Graphs (DOGs) in seconds for a real-world use case. The results? A single, seamless process handling:

✅ Transactional Steps – Actions like bookings
Analytical Steps – AI-powered pricing and decisioning

This is a huge leap for AI-driven business transformation.

The End of the Operational vs. Analytical Divide

Historically, operational data systems (handling real-time transactions) and analytical systems (for decision-making, forecasting, and reporting) have been separate worlds. This led to:

Lag Between Insight & Action – Analytics were always a step behind.
Complex Integration Efforts – Keeping the two in sync was a nightmare.
Higher Costs & Slower Transformation – Running dual infrastructures added complexity and delays.

But with Agentic Data Object Graphs (DOGs), we can now combine both in a single, AI-driven framework—no more silos.

AI Digital Twins: The New Paradigm

With DOGs, businesses can model both operational and analytical processes as a single unified execution graph.

  • Immediate Decision-Making – No waiting for analytics to catch up.
  • Adaptive Processes – AI dynamically refines transactions based on real-time insights.
  • E2E Transformation – True business transformation, where AI, data, and process automation work together seamlessly.

This is not just another tech evolution—it’s a fundamental shift in how business systems operate.

Why This Matters Right Now

We’re entering a new era where data, AI, and business operations are no longer separate conversations.

With AI Digital Twins and Agentic Data Object Graphs, organizations can finally:

  • Run real-time operations and AI-driven decision-making together
  • Optimize both transactional and analytical flows simultaneously
  • Reduce complexity, cost, and technical debt

It’s an amazing time to be at the nexus of Business, Tech, and AI, driving true end-to-end transformation.

If you’re still treating operational and analytical processes separately—it might be time to rethink that approach.

🚀 The future isn’t coming—it’s already here.

18. AI Breeds Better DOG

Proof That the Operational-Analytical Divide Is Disappearing 🚀

At Dataception, we’ve been working on something big—getting AI to generate Data Object Graphs (DOGs) for business processes as AI Digital Twins.

From Process Design to Machine Execution

One of the biggest challenges in business transformation is bridging the gap between process design and execution. Traditionally, business processes exist as static diagrams, documents, and flowcharts—but translating those into real, working solutions has always been a manual and time-consuming effort.

With Agentic DOGs, we automate this entire process by turning traditional business workflows into machine-executable data object graphs, where each node represents a concrete data product—whether it’s:
✅ A metric
✅ A dataset
✅ An ML or GenAI model
✅ A decision engine

Real-World Example: Customer Lifetime Value as a DOG

We recently mapped customer lifetime value (CLV) as a Data Object Graph. Here’s what happened:

  1. Each step automatically extracted and processed the precise data needed—pulling from sources like Zoho and other platforms.
  2. The data seamlessly flowed through interconnected data products, linking metrics, models, and decision logic.
  3. GenAI didn’t just generate the process—it became an active participant, analyzing data and generating insights dynamically along the way.

Why This Is a Game-Changer

🚀 No more manually crafting business process diagrams—everything is automated and executable.
🚀 Cross-domain processes become easier to maintain, as AI dynamically adapts to changing business needs.
🚀 Business process transformation moves from months to days, with real-time AI-driven iteration.

The Future of AI-Driven Business Execution

By using AI-generated Data Object Graphs, we’re not just automating business processes—we’re creating adaptive, intelligent AI Digital Twins that:

🔹 Bridge the gap between design and execution
🔹 Enable AI to actively participate in decision-making
🔹 Dynamically integrate with real business data in real-time

This is the next step in AI-powered business transformation, and we’re seeing incredible results already.

If you’re working on similar challenges—or just want to see this in action—let’s connect.

With  Dataception's DOGs (Data Object Graphs), AI is just a walk in the park. 🐕🚀

Want to Learn More About Our DOGs (Data Object Graphs)?

Get in touch to see how we can ​revolutionize the way your businesses handles data and analytics through our Rapid Predictable Data and AI Delivery model.

Explore our Services​  Contact Us Today​