Engram AI Periodic Table Business Plan

Elena Rivera’s strategic mapping of Engram capabilities to Martin Keen’s AI Periodic Table, demonstrating competitive strength in each element.

Executive Summary

Martin Keen’s AI Periodic Table provides a framework for understanding the building blocks of modern AI systems. This business plan maps Engram’s architecture to each element, identifying:

Strong: Production-ready implementation
Emerging: Partially implemented, roadmap priority
Gap: Not yet addressed, opportunity for development

Key Finding: Engram demonstrates Strong implementation in 12 of 18 elements, with our proposed Gk (Graph Knowledge) element representing a unique competitive advantage.

The AI Periodic Table Structure

Rows (Maturity Levels)

Row	Name	Description
R1	Primitives	Foundational building blocks
R2	Compositions	Combined capabilities
R3	Deployment	Production patterns
R4	Emerging	Cutting-edge innovations

Columns (Capability Domains)

Column	Name	Description
C1	Reactive	Response to stimuli
C2	Retrieval	Information access
C3	Orchestration	Workflow coordination
C4	Validation	Quality assurance
C5	Models	AI model types

Element-by-Element Analysis

Row 1: Primitives

Pr (Prompts) — C1 Reactive × R1 Primitives

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Agent system prompts
Location	`backend/agents/{elena,marcus,sage}.py`
Strength	Three distinct agent personalities with rich, contextual system prompts
Differentiation	Prompts are not static—they incorporate tri-search context dynamically

Business Value: Engram’s prompts define personality, not just function. Elena, Marcus, and Sage each have distinct voices, expertise areas, and behavioral patterns that create a memorable user experience.

Em (Embeddings) — C2 Retrieval × R1 Primitives

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Zep embedding layer
Location	`backend/memory/client.py`
Strength	Automatic embedding of all conversations and artifacts
Model	OpenAI ada-002 via Zep Cloud

Business Value: Every interaction is automatically embedded and searchable. No manual indexing required—memory builds organically through use.

Lg (LLM) — C5 Models × R1 Primitives

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Multi-model orchestration
Location	`backend/llm/{claude,gemini}_client.py`, model router
Strength	Claude Sonnet 4, Gemini 2.0 Flash, GPT-4o, with automatic fallback
Architecture	Model router selects optimal model per query type

Business Value: Not locked to a single vendor. Engram uses the best model for each task while maintaining resilience through fallback chains.

Row 2: Compositions

Fc (Function Call) — C1 Reactive × R2 Compositions

Status: 🟢 Strong

Aspect	Engram Implementation
Component	MCP (Model Context Protocol) Tools
Location	`backend/api/routers/mcp.py`, `mcp_server.py`
Strength	40+ tools including memory, search, delegation, schema management
Protocol	MCP-compatible, works with Claude Desktop and other MCP clients

Business Value: Engram agents can take action, not just answer questions. They query databases, create stories, search memory, and delegate tasks.

Vx (Vector) — C2 Retrieval × R2 Compositions

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Zep vector store
Location	Memory client semantic search
Strength	Tri-Search: semantic (vector) + keyword + graph combined
Scale	Cloud-hosted, handles millions of embeddings

Business Value: Tri-Search surpasses traditional RAG by combining three retrieval methods. Users get better context without knowing how search works.

Rg (RAG) — C3 Orchestration × R2 Compositions

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Context assembly pipeline
Location	`backend/agents/base.py`, memory enrichment
Strength	Automatic retrieval-augmented generation on every query
Architecture	Search → Filter → Inject → Generate → Store

Business Value: Every agent response is grounded in memory. Hallucinations reduced through automatic context injection.

Gr (Guardrails) — C4 Validation × R2 Compositions

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Enterprise security middleware
Location	`backend/api/middleware/auth.py`
Strength	Azure Entra ID integration, tenant isolation, RBAC
Compliance	Enterprise-ready authentication for all endpoints

Business Value: Not a toy—Engram is built for enterprise from day one. Multi-tenant isolation ensures customer data sovereignty.

Mm (Multimodal) — C5 Models × R2 Compositions

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Image + Voice + Text
Location	`gemini_client.py` (Imagen 3.0), `voice.py` (VoiceLive)
Strength	Story visuals via Imagen, real-time voice via Azure VoiceLive
Architecture	Two-step image generation: spec → image

Business Value: Engram creates, not just answers. Stories include visuals. Voice enables hands-free interaction.

Row 3: Deployment

Ag (Agent) — C1 Reactive × R3 Deployment

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Elena, Marcus, Sage
Location	`backend/agents/`
Strength	Three production agents with distinct capabilities
Specialization	Elena (orchestrator), Marcus (engineer), Sage (storyteller)

Business Value: Specialized agents outperform generalist chatbots. Users get expert assistance matched to their need.

Ft (Finetune) — C2 Retrieval × R3 Deployment

Status: 🟡 Gap

Aspect	Status
Current	Not implemented
Roadmap	Custom fine-tuned models for domain-specific tasks
Opportunity	Fine-tune Granite or Llama for Engram-specific patterns

Business Value: Future opportunity for specialized, lower-cost inference on common tasks.

Fw (Framework) — C3 Orchestration × R3 Deployment

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Temporal workflows
Location	`backend/workflows/`
Strength	Durable execution for multi-step agent tasks
Architecture	Story, agent, conversation workflows with retry policies

Business Value: Engram doesn’t fail silently. Long-running tasks survive outages and can be monitored in real-time.

Rt (Red-team) — C4 Validation × R3 Deployment

Status: 🟡 Emerging

Aspect	Status
Current	Basic input validation
Roadmap	Adversarial testing, prompt injection detection
Opportunity	Integration with Azure AI Content Safety

Business Value: Critical for enterprise trust. Roadmap item for security-conscious customers.

Sm (Small) — C5 Models × R3 Deployment

Status: 🟡 Gap

Aspect	Status
Current	Not implemented
Roadmap	IBM Granite 3.3, Microsoft Phi-4 for edge deployment
Opportunity	Lower costs, faster inference, on-premises option

Business Value: Small models enable cost-effective high-volume scenarios and edge deployment without cloud dependency.

Row 4: Emerging

Ma (Multi-agent) — C1 Reactive × R4 Emerging

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Agent delegation
Location	`backend/agents/delegation.py`
Strength	Elena delegates to Sage for stories, to Marcus for technical tasks
Architecture	Delegation via MCP tools, results flow back to originator

Business Value: Complex tasks are automatically routed to specialists. Users don’t need to know which agent to ask.

Sy (Synthetic) — C2 Retrieval × R4 Emerging

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Story + visual generation
Location	`story_workflow.py`, `story_activities.py`
Strength	Claude generates narratives, Gemini generates visuals
Output	Markdown stories with embedded architecture diagrams and images

Business Value: Engram creates knowledge, not just retrieves it. Technical concepts become memorable stories with visuals.

Gk (Graph Knowledge) — C3 Orchestration × R4 Emerging

Status: 🟢 UNIQUE STRENGTH ⭐

Aspect	Engram Implementation
Component	Zep temporal knowledge graph
Location	`backend/memory/client.py`, Zep Cloud integration
Strength	Entity extraction, fact linking, temporal awareness
Architecture	Knowledge compounds across sessions automatically

Capabilities:

Entity Extraction: People, systems, concepts extracted from every conversation
Fact Linking: Relationships stored as graph edges with timestamps
Temporal Queries: “What did Derek know about HorizonDB last week?”
Cross-Session Learning: Knowledge persists and grows without explicit ingestion

Why This Is Our Moat:

While competitors use static RAG (retrieve → augment → generate), Engram uses dynamic Graph Knowledge orchestration. Context assembly is driven by semantic understanding of entity relationships, not just vector similarity.

Business Value: Engram remembers relationships, not just facts. This enables true continuity across conversations and compound learning over time.

In (Interpret) — C4 Validation × R4 Emerging

Status: 🟡 Gap

Aspect	Status
Current	Not implemented
Roadmap	Explainability for agent decisions
Opportunity	“Why did Elena suggest this?” transparency

Business Value: Critical for regulated industries. Enables audit trails for AI-assisted decisions.

Th (Thinking) — C5 Models × R4 Emerging

Status: 🟢 Strong

Aspect	Engram Implementation
Component	Extended thinking in workflows
Location	Temporal activity design
Strength	Multi-step reasoning with query handlers for progress
Architecture	Workflows support long-running “thinking” tasks

Business Value: Complex reasoning isn’t rushed. Story generation takes 2-3 minutes because quality matters more than speed.

Competitive Position Summary

Status	Count	Elements
🟢 Strong	12	Pr, Em, Lg, Fc, Vx, Rg, Gr, Mm, Ag, Fw, Ma, Sy, Gk, Th
🟡 Emerging/Gap	4	Ft, Rt, Sm, In

Unique Differentiator: Gk (Graph Knowledge)

Engram is the only platform with production-ready Graph Knowledge implementation. While:

OpenAI offers GPT-4 (Lg) and function calling (Fc)
Anthropic offers Claude (Lg) and thinking (Th)
Microsoft offers Azure AI (Lg) with guardrails (Gr)

Only Engram combines all of these with temporal knowledge graphs that enable true context engineering.

Roadmap Priorities

Sm (Small Models): Deploy Phi-4 or Granite for high-volume edge scenarios
In (Interpret): Build explainability layer for enterprise compliance
Rt (Red-team): Integrate adversarial testing for security assurance
Ft (Finetune): Custom models for domain-specific tasks

Visual: Engram on the AI Periodic Table

     C1 Reactive   C2 Retrieval   C3 Orchestration   C4 Validation   C5 Models
   ┌──────────────┬──────────────┬──────────────────┬───────────────┬──────────────┐
R1 │   🟢 Pr      │   🟢 Em      │                  │               │   🟢 Lg      │
   │   Prompts    │  Embeddings  │                  │               │    LLM       │
   ├──────────────┼──────────────┼──────────────────┼───────────────┼──────────────┤
R2 │   🟢 Fc      │   🟢 Vx      │     🟢 Rg        │    🟢 Gr      │   🟢 Mm      │
   │ Function Call│   Vector     │      RAG         │  Guardrails   │  Multimodal  │
   ├──────────────┼──────────────┼──────────────────┼───────────────┼──────────────┤
R3 │   🟢 Ag      │   🟡 Ft      │     🟢 Fw        │    🟡 Rt      │   🟡 Sm      │
   │    Agent     │  Finetune    │   Framework      │  Red-team     │    Small     │
   ├──────────────┼──────────────┼──────────────────┼───────────────┼──────────────┤
R4 │   🟢 Ma      │   🟢 Sy      │     ⭐ Gk        │    🟡 In      │   🟢 Th      │
   │ Multi-agent  │  Synthetic   │ Graph Knowledge  │  Interpret    │  Thinking    │
   └──────────────┴──────────────┴──────────────────┴───────────────┴──────────────┘

🟢 = Strong    🟡 = Gap/Emerging    ⭐ = Unique Differentiator

This business plan was prepared by Elena Rivera, Chief Strategy Officer for Engram.