Signals

Foundational Layer

Enterprises faced a build-vs-buy decision almost overnight. The arrival of LLMs forced organizations to rethink their entire data strategy — what data they have, how clean it is, and whether it's usable for fine-tuning. Open-source models (Llama, Mistral, etc.) then complicated the calculus further — suddenly self-hosting became viable, raising questions about data sovereignty, cost control, and avoiding vendor lock-in.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Large Language Models (LLMs) - Neural network models trained on large corpora of text and code to predict and generate language. LLMs serve as the foundational reasoning and generation layer for modern AI applications, enabling tasks like summarization, translation, planning, and code synthesis.
  • Generative Pre-trained Transformer (GPT) - A class of transformer-based language models pre-trained on broad datasets and fine-tuned for specific tasks. GPTs are designed to generate coherent, context-aware text and are commonly used as general-purpose AI engines.
  • Open-Source LLMs - Language models whose weights, architectures, or training code are publicly available. They enable self-hosting, customization, transparency, and ecosystem innovation outside proprietary platforms.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Artificial Intelligence - Measuring the AI investment occurring from ChatGPT usage to MCP to investing in agentic automation, evaluating a company's grasp of it.
  • Cloud - Measuring the cloud investment, beginning with which clouds they use, but then looking at their approach to managing the technical and business side.
  • Open-Source - Measuring the open-source investment, and how much open-source they use, but also potentially contribute to, and even if they are investing in inner source.
  • Languages - Which programming languages are used by teams, understanding the diversity of languages in use, and the relationship to services and tooling.
  • Code - Measuring the code investment, and what libraries and frameworks are in use, as well as any software development kits that are provided or being applied for integrations.

Every enterprise now needs an AI infrastructure opinion, whether they like it or not. The foundation you choose — proprietary, open-source, or hybrid — sets the constraints for everything built on top of it.

Retrieval & Grounding

These emerged because raw LLMs hallucinate and lack enterprise context. Vector databases became a new infrastructure category teams had to evaluate and operate. RAG became the de facto pattern for grounding models in proprietary data without expensive fine-tuning. Prompt engineering briefly looked like a new discipline but increasingly gets absorbed into software engineering. As RAG, memory, tool results, and multi-turn conversation history all compete for space in finite context windows, context engineering has emerged as a distinct discipline. It's no longer just about writing a good prompt — it's about orchestrating what the model sees — which retrieved documents make the cut, how conversation history gets summarized, what metadata gets injected, and how tool outputs are formatted.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Vector Databases - Datastores optimized for storing and querying high-dimensional embeddings. They enable semantic search, similarity matching, and contextual retrieval by comparing vector representations rather than exact keywords.
  • Retrieval-Augmented Generation (RAG) - An architectural pattern that combines external data retrieval with language model generation. Retrieved context is injected into prompts to ground responses in up-to-date or domain-specific information.
  • Prompt Engineering - The practice of designing inputs, instructions, and examples to guide model behavior. Prompt engineering shapes outputs without changing model weights, acting as a lightweight control layer over model capabilities.
  • Context Engineering - The discipline of designing and managing the full context provided to a model — including system prompts, retrieved documents, tool outputs, conversation history, memory, and structured metadata. Context engineering evolves beyond prompt engineering into a holistic practice of shaping everything a model sees before it generates a response.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Data - Measuring the data investment, and how strong the data teams are, and what are they focused on from access, quality, analytics, to governance and compliance issues.
  • Databases - Measuring the database investment, and what database platforms are in use, and what database tooling is in use across teams to provide data access.
  • Virtualization - Measuring the virtualization investment including data, examples, synthetic data, but also API mocking, and other ways companies are virtualizing resources.
  • Specifications - Measuring the specifications in use, such as OpenAPI, AsyncAPI, and JSON Schema, but also newer formats like A2A, MCP, and other AI specs.
  • Context Engineering - Measuring the investment in context assembly and management — context window optimization strategies, context prioritization frameworks, summarization pipelines for conversation history, metadata injection patterns, and the tooling used to compose, test, and monitor what models actually see at inference time.

Organizations realized they need robust knowledge management and data pipelines before AI adds value — exposing years of neglected data hygiene — and they need deliberate strategies for context assembly and prioritization, because what you don't put in the context window matters as much as what you do.

Customization & Adaptation

Sitting between the foundation layer and retrieval, enterprises discovered that prompt engineering and RAG aren't always enough. Fine-tuning on proprietary data — customer interactions, internal documents, domain-specific corpora — became a real investment area, especially in regulated industries like healthcare, legal, and finance where general-purpose models lack the precision required. Multimodal capabilities expanded the surface area further — claims processing from images, quality inspection from video, document extraction from PDFs, and meeting summarization from audio.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Fine-Tuning & Model Customization - The process of further training pre-trained models on domain-specific or proprietary data to improve performance on targeted tasks. Fine-tuning sits between prompt engineering and training from scratch, offering deeper behavioral customization while inheriting general capabilities from the base model.
  • Multimodal AI - Models and systems capable of processing and generating across multiple data types — text, images, audio, video, and documents. Multimodal capabilities extend AI beyond language-only tasks into visual inspection, document understanding, meeting transcription, and cross-modal reasoning.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Data Pipelines - Measuring investment in training and fine-tuning data pipelines — how organizations curate, label, version, and govern the proprietary datasets used to customize models, including text, image, audio, and video corpora.
  • Model Registry & Versioning - Measuring whether enterprises are tracking which models (base, fine-tuned, adapted) are deployed where, including version lineage, performance baselines, and rollback capabilities.
  • Multimodal Infrastructure - Measuring the investment in processing non-text data — document extraction (OCR, PDF parsing), image and video analysis, audio transcription, and the pipelines that normalize these inputs for model consumption.
  • Domain Specialization - Measuring the degree to which organizations are building or procuring domain-specific models versus relying on general-purpose models, and the regulatory or compliance drivers behind that choice.

Enterprises now need data pipelines that handle not just text but images, audio, and video — and the fine-tuning infrastructure to adapt models to their specific domain without exposing proprietary data to third parties.

Efficiency & Specialization

As costs and latency became real concerns, enterprises learned that not every task needs the biggest model. Small language models allow edge deployment and lower inference costs. Model routing lets systems dynamically choose the right model for the right task, optimizing cost-quality tradeoffs. Reasoning models (like o1-style chains of thought) opened up tasks previously too complex for LLMs — planning, multi-step analysis, code verification.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Small Language Models (SLMs) - Compact language models optimized for efficiency, speed, and on-device or edge deployment. SLMs trade scale for controllability and cost, making them suitable for constrained or embedded environments.
  • Model Routing / Orchestration - Systems that dynamically select, sequence, or combine multiple models and tools to fulfill a task. Routing optimizes for cost, latency, accuracy, or capability by choosing the right model at the right time.
  • Reasoning Models - Models explicitly optimized to perform multi-step reasoning, planning, and problem decomposition. They emphasize structured thought over raw text generation to improve correctness on complex tasks.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Automation - Measuring the automation investment in all of its forms to understand how sophisticated automation is, and how much it is being applied across operations.
  • Containers - Measuring the container investment, beginning with Docker, but moving to the cloud, and where Kubernetes is in their overall platform journey with containers.
  • Platform - Measuring the platform investment, and where a company is at in their platform journey, evaluating what common services, guard rails, and roles are in place.
  • Operations - Measuring the operational investment, and how much they think about the big picture strategy of their operations, and how they can be improving.

Enterprises now need an inference orchestration layer — a genuinely new piece of architecture most organizations didn't have a year ago — to match workloads to the right model at the right cost.

Productivity

These are the most visible enterprise touchpoints. They force organizations to confront questions around code IP, security review of AI-generated code, developer productivity measurement, and license compliance. Copilots extend beyond code into sales, support, HR, and finance — meaning every business function starts asking for "their copilot."

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Coding Assistants - AI tools embedded in development environments that assist with writing, refactoring, debugging, and understanding code. They translate natural language intent into executable code and provide real-time developer feedback.
  • Copilots - AI assistants designed to work alongside humans within specific workflows. Copilots provide suggestions, automation, and contextual assistance while keeping humans in the decision loop.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Software As A Service (SaaS) - Measuring the SaaS investment when it comes to optimization, FinOps, and other areas, to understand how much investment is in this area.
  • Code - Measuring the code investment, and what libraries and frameworks are in use, as well as any software development kits that are provided or being applied for integrations.
  • Services - The entire SaaS portfolio for companies, beginning with the number of services, but then also evaluating which are infrastructure, platform, or more business.

The pressure on IT to standardize rather than let a thousand tools bloom is real — without deliberate platform choices, enterprises risk a fragmented landscape of disconnected AI assistants with inconsistent security postures.

Integration & Interoperability

MCP (Model Context Protocol) and the broader agent wave represent perhaps the most architecturally disruptive shift. Agents that can take actions — not just generate text — mean enterprises need to think about authorization, audit trails, sandboxing, and failure modes in entirely new ways. MCP specifically aims to standardize how models connect to tools and data sources, which has huge implications for reducing integration fragmentation.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • MCP (Model Context Protocol) - A protocol for exposing tools, APIs, and data sources to models in a structured, machine-readable way. MCP standardizes how models discover, invoke, and reason about external capabilities.
  • Agents - Autonomous or semi-autonomous systems that use models, tools, and memory to pursue goals over time. Agents can plan, act, observe outcomes, and iterate without continuous human prompting.
  • Skills - Discrete, reusable capabilities exposed to models or agents as callable functions. Skills encapsulate business logic, integrations, or workflows that extend model usefulness beyond text generation.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • API - Measuring the overall API investment, from being API-first to design-first, to full lifecycle API management to understand where they are in their API journey.
  • Integrations - Measuring the integration investment involving iPaaS, embedded iPaaS, but also legacy approaches with ETL, batch, and other common ways of integrating.
  • Event-Driven - Measuring the event-driven investment, and looking at the types of APIs in use, and the technology they are using that is steering them towards event-driven.
  • Patterns - Measuring the different patterns in use across the different types of APIs, but also the parts and pieces of integrations, to understand the diversity of patterns.
  • Specifications - Measuring the specifications in use, such as OpenAPI, AsyncAPI, and JSON Schema, but also newer formats like A2A, MCP, and other AI specs.
  • Apache - Measuring the Apache tooling investment, and what projects are in use, and how they are leveraged as part of operations, including involvement in community.
  • CNCF - Measuring the CNCF tooling investment, and what projects are in use, and how they are being leveraged as part of operations, including involvement in community.

Skills — modular capabilities agents can invoke — push toward composable AI systems. The organizations that get integration right will compound their AI capabilities; those that don't will drown in bespoke point-to-point connections.

Statefulness

Memory introduces persistent context across sessions, which means AI systems can learn user preferences, accumulate project context, and behave more like collaborators than stateless tools.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Memory Systems - Mechanisms for storing, retrieving, and updating information across interactions. Memory systems enable personalization, long-running tasks, and continuity beyond a single prompt or session.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Observability - Measuring the state of observability, how they are monitoring, testing, tracing, and reporting on their operations via dashboards, and other approaches.
  • Governance - Measuring the governance that is occurring, and how focused it is on APIs, as well as aligned with wider security, compliance, and other aspects of governance.
  • Security - Measuring the security investment, and whether or not it is still more application focused or has evolved to be more API-centered, as well as thinking about AI.
  • Data - Measuring the data investment, and how strong the data teams are, and what are they focused on from access, quality, analytics, to governance and compliance issues.

For enterprises, this raises serious questions about data retention policies, privacy compliance (GDPR right to deletion, for instance), and the blurring line between "tool" and "system of record." It also creates genuine competitive advantage for organizations that implement it well — their AI gets better over time in ways competitors can't easily replicate.

Measurement & Accountability

Without systematic evaluation, enterprises are investing in AI with no reliable feedback loop. This wave encompasses model-level benchmarks, but more critically, enterprise-specific evaluation — Is the RAG pipeline returning accurate results? Are agents completing tasks successfully? How often do models hallucinate in production? Evaluation is becoming its own discipline — evals as code, continuous benchmarking against regression, A/B testing of prompt strategies, and human-in-the-loop scoring.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Evaluation & Benchmarking - The practice of systematically measuring model and system performance across accuracy, reliability, safety, and task completion. Evaluation frameworks include model-level benchmarks, RAG quality scoring, agent success rates, hallucination detection, and enterprise-specific acceptance criteria.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Testing & Quality - Measuring the investment in AI-specific testing — eval frameworks, regression benchmarks, hallucination detection, RAG accuracy scoring, and agent task completion rates as part of CI/CD and production monitoring.
  • Observability - Measuring whether observability extends beyond traditional infrastructure metrics into model-level telemetry — latency per model, token usage, confidence scores, retrieval quality, and drift detection.
  • Developer Experience - Measuring how enterprises are instrumenting AI developer workflows — adoption metrics for coding assistants, productivity baselines, internal satisfaction surveys, and the feedback loops between developers and AI platform teams.
  • ROI & Business Metrics - Measuring whether organizations have connected AI system performance to business outcomes — time saved, error reduction, customer satisfaction, cost avoidance — or whether measurement remains purely technical.

Organizations that treat evaluation as an afterthought will continue to ship unreliable AI, while those that invest in measurement infrastructure will compound quality improvements over time. Measurement is the difference between AI as an experiment and AI as a reliable capability.

Governance & Risk

The EU AI Act, executive orders, and emerging corporate AI policies represent a wave that is reshaping what gets built and how. Enterprises are standing up AI review boards, model registries, and risk classification systems. This isn't just compliance theater — it determines which use cases proceed to production and which get blocked.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Governance & Compliance - Organizational and regulatory frameworks for managing AI risk, accountability, and transparency. This includes model registries, risk classification, bias auditing, explainability requirements, and adherence to emerging regulation like the EU AI Act and executive orders.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Regulatory Posture - Measuring how enterprises are responding to AI regulation — EU AI Act classification, risk assessments, model documentation, and whether compliance is proactive or reactive.
  • AI Review & Approval - Measuring whether formal AI review processes exist — review boards, use case approval workflows, model risk tiering, and the speed at which new AI use cases move from proposal to production.
  • Security - Measuring whether security practices have extended to AI-specific threats — prompt injection, model extraction, training data poisoning, and the authorization boundaries around autonomous agents.
  • Governance - Measuring the governance that is occurring, and how focused it is on APIs, as well as aligned with wider security, compliance, and other aspects of governance.
  • Privacy & Data Rights - Measuring investment in AI-specific privacy infrastructure — consent management for training data, right-to-deletion compliance across memory systems, data lineage tracking, and cross-border data flow management for model training and inference.

Governance intersects with every other wave — foundation model selection is constrained by regulatory posture, agent autonomy is bounded by risk tolerance, and memory systems must comply with data rights. Governance is not a layer on top — it's a filter that runs through every architectural decision.

Economics & Sustainability

AI inference costs are surprising enterprises the way cloud bills did a decade ago, but with less predictability. Token-level pricing, GPU procurement and availability, the cost differential between proprietary and open-source hosting, and the hidden costs of fine-tuning all create a new financial management challenge. Dependency on a small number of model providers and chip manufacturers introduces supply chain concentration risk that most enterprises haven't experienced in software before. Data center capacity — owned, leased, and cloud-provisioned — has become a strategic constraint. GPU availability, power costs, cooling infrastructure, and geographic placement for data sovereignty all factor into where and how enterprises can run AI workloads.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Cost Economics & FinOps - The discipline of understanding, forecasting, and optimizing the financial cost of AI operations — including inference spend, GPU procurement, token-level pricing, training costs, and model hosting. AI FinOps applies cloud cost management principles to a new and often less predictable cost surface.
  • Supply Chain & Dependency Risk - The set of dependencies enterprises take on when building with AI — model providers, chip manufacturers, API pricing stability, open-source model licensing, and the risk of deprecation or breaking changes. Supply chain risk in AI mirrors software supply chain concerns but with less mature tooling and higher concentration among a small number of providers.
  • Data Centers - The physical and cloud infrastructure housing the compute required for AI training and inference. Data centers encompass GPU clusters, power and cooling capacity, geographic placement for data sovereignty, and the massive capital expenditure underpinning AI capabilities. They are the material foundation that every other wave depends on.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • AI FinOps - Measuring the maturity of AI cost management — token-level cost tracking, inference spend forecasting, model cost-performance optimization, GPU utilization monitoring, and chargeback models for shared AI infrastructure.
  • Provider Strategy - Measuring the deliberateness of model provider and infrastructure choices — single vs. multi-provider strategies, contractual terms, switching costs, and the balance between proprietary APIs and self-hosted open-source alternatives.
  • Partnerships & Ecosystem - Measuring the strategic AI partnerships announced and in practice — cloud AI partnerships (Azure OpenAI, AWS Bedrock, GCP Vertex), model provider relationships (OpenAI, Anthropic, Cohere, Mistral), and how these partnerships shape or constrain architectural choices.
  • Talent & Organizational Design - Measuring how enterprises are staffing AI initiatives — new roles (ML platform engineer, AI product manager, prompt engineer), team structures (centralized AI teams vs. embedded), and the skills gaps that job postings reveal about organizational readiness.
  • Data Centers - Measuring the data center investment and strategy — owned vs. leased vs. cloud GPU capacity, geographic distribution for latency and sovereignty, power and cooling infrastructure, capital expenditure commitments, and how data center constraints shape model selection and deployment architecture.

AI infrastructure planning now extends into physical-world decisions that most software organizations haven't had to make since the pre-cloud era, and the capital intensity of data centers is reshaping build-vs-buy calculations at the hardware layer. Enterprises need AI-specific FinOps practices and deliberate supply chain diversification strategies — choosing which providers to depend on is now a strategic decision with real downside risk.

Storytelling & Entertainment & Theater

Develop processes for gathering and processing when it comes to the theater and entertainment from the artificial intelligence industry. Identifying what is entertainment and theater in an open and safe way, encouraging teams to think about what is possible, but understand where it fits into the bigger landscape of the enterprise.

• Waves - These are the waves that have resulted in a shift across the enterprise today.
  • Moltbook - A notebook-centric development pattern where code, prompts, data, and execution co-evolve. Moltbooks emphasize rapid iteration and mutation, often blurring the line between experimentation and production.
  • Gastown - A metaphor for early-stage AI infrastructure: dense, experimental, and rapidly evolving. Gastown systems prioritize velocity and exploration over polish, governance, or long-term maintainability.
  • Ralph Wiggum - A shorthand for AI systems that are earnest but unreliable — confidently producing outputs without true understanding. The term highlights failure modes where models appear capable but lack grounding or judgment.
  • OpenClaw / Clawdbot - Experimental, community-driven AI tools and bots that emerge from hackathons, side projects, and open experimentation. They represent the playful, exploratory edge of AI development — often more theater than production — but occasionally surface patterns or capabilities that inform more serious enterprise work.
  • Artificial General Intelligence (AGI) - A hypothetical form of AI capable of understanding, learning, and applying knowledge across a wide range of tasks at a human or superhuman level. AGI implies general reasoning ability rather than task-specific competence.
• Signals - These are the signals Naftiko is tuned into to understand where companies are.
  • Alignment - Measuring the business alignment investment, and are they doing work to bridge engineering with business, and invest more into the productization of APIs.
  • Standardization - Measuring the standardization investment, beginning with what standards they intentionally or unintentionally use, but also their strategic approach.
  • Mergers & Acquisitions - Measuring how many mergers and acquisitions are conducted, and how their operations are shaped by years of this M&A approach to innovation.
  • Experimentation & Prototyping - Measuring the investment in experimental AI tools and community-driven projects — hackathon outputs, internal bot experimentation (OpenClaw / Clawdbot patterns), prototype-to-production pipelines, and how organizations create safe spaces for exploratory AI work that informs but doesn't disrupt production systems.

Properly mapping any new technology to the existing business landscape, providing solutions that are attached to real world outcomes, while still celebrating and enjoying what is happening across the technology spectrum.