Funnel Amp’s Mission-Aligned AI Standard (MAAS)

Purpose and Scope

Published June 16, 2026

Funnel Amp is an AI Knowledge Base Architecture and RAG Documentation practice serving B2B organizations in Green Energy and Sustainable Finance. Its core mission is to assist businesses that take an environmentally responsible and humanity-first position as they adopt, develop, and deploy AI solutions and systems.

Funnel Amp’s ultimate goal when assisting a client, is ensuring that an AI project, or the component directly under Funnel Amp’s control, is implemented with a minimized carbon footprint and maximized efficiency while meeting or exceeding a client’s expected outcomes.

This MAAS is a living framework organized around eight foundational elements that govern how Funnel Amp engages AI technology. It’s designed to be reviewed and updated as the AI landscape, applicable regulations, and Funnel Amp’s own capabilities evolve.

Elements Status Key

Active

Fully Implemented

In Progress

Actively Underway

Planned

Future Action

The Eight Elements

  1. Justification Analysis

Active

Before initiating any AI-assisted workflow within its own operations, or recommending AI deployment to a client, Funnel Amp documents a clear business case establishing that AI is the appropriate tool for the problem. This analysis asks whether the operational benefit of deploying AI (measured in efficiency gains, accuracy improvements, or mission outcomes) demonstrably outweighs any associated environmental or human costs.

In Funnel Amp’s own practice, this standard is applied to every tool and workflow decision:

  • Client RAG builds are scoped after a Knowledge Base Audit confirms that AI retrieval will produce a measurable improvement over existing processes.
  • Tool selection during builds is evidence-based, not default-based. Embedding models, chunking strategies, and retrieval configurations are selected only after documented corpus analysis demonstrates they are the appropriate choice for that specific dataset.
  • AI is not invoked for tasks that can be completed reliably and efficiently by a human workflow. This prevents unnecessary compute consumption and preserves the credibility of AI-assisted outputs.
  1. Human-in-the-Loop (HITL) Design

Active

Funnel Amp incorporates Human-in-the-Loop design into both its own workflows, and the knowledge base architectures it builds for clients. HITL is treated as a dual mandate: it’s simultaneously an accountability mechanism that ensures consequential AI outputs are reviewed by a qualified person before triggering action, and an environmental mechanism that prevents the unnecessary compute consumption generated by unsupervised agentic systems.

In Funnel Amp’s internal operations, HITL design is applied in the following manner:

  • AI functions as a junior research assistant, analysis, co-strategist or co-developer. Sequoia Brown (Funnel Amp’s lead developer) exercises senior authority over approving, editing and finalizing all AI output. No AI output is integrated into a workflow or handed off to a client without human review.

In client knowledge base architecture, HITL plays the following role:

  • HITL design principles are built into the retrieval architecture recommendations Funnel Amp delivers. Clients are advised on where human review checkpoints should be embedded in their agentic workflows, particularly at decision points that affect regulatory compliance, safety protocols, or public-facing outputs.
  1. Clear Human Accountability

Active

Every AI deployment within Funnel Amp’s operations has assigned human ownership over outputs and outcomes. AI automation is not approached as a “set it and forget it” default option. If automation is employed it is used sparingly and with clearly assigned human responsibility. As Funnel Amp’s founder and Senior developer, Sequoia Brown is the accountable owner of every AI-assisted decision, every deliverable, and every client-facing output produced under the Funnel Amp name.

Within Funnel Amp, clearly assigned human accountability ensures that:

  • There is no diffusion of responsibility across projects, operations, teams, or departments.
  • Every AI output that affects a client’s operations is the direct professional responsibility of a named individual.
  • Clients engaging Funnel Amp know precisely who is accountable for the quality and integrity of the knowledge architecture delivered to them.

For clients, Funnel Amp makes the following recommendation:

  • Organizations designate a named Knowledge Base Owner responsible for governance, update schedules, and retrieval quality oversight, consistent with the accountability structure outlined in this MAAS.
  1. Knowledge Base Quality and Governance Policy

Active

The quality of a knowledge base and the RAG documentation layer directly impacts the reliability and accuracy of an AI agent. Funnel Amp’s entire practice is built around the principle that these crucial components should be crafted with the highest level of precision and care possible. The following standards govern every knowledge base and RAG layer that Funnel Amp builds, audits, or refreshes:

Funnel Amp’s practice is guided by the following development framework:

  • Clean Corpus Protocol — Noise and non-retrievable content removed before any embedding occurs.
  • Evidence-Based Architecture Selection — Chunking and embedding choices validated by corpus analysis, not assumption.
  • Two-Tier Metadata Schema — Versioned document and chunk-level fields enabling precision filtering and retrieval.
  • Controlled Vocabulary Architecture — Domain-specific tags derived directly from source documents for semantic accuracy.
  • Retrieval Precision Testing — Every knowledge base evaluated against a purpose-built golden dataset before delivery.
  • Knowledge Decay Governance — Structured review triggers that prevent stale content from becoming a liability.
  1. Renewable Energy and SLM Infrastructure

In Progress

Funnel Amp is committed to keeping the computational and carbon footprint of its operations as minimal as possible. This commitment is reflected in its current operational and infrastructure choices, and planned action items on its carbon minimization road map.

Here are Funnel Amp’s current practices and planned action items:

  • All client RAG builds are conducted entirely on local hardware (Ryzen 9 9900X, 64GB RAM, AMD RX 7600, Pop!_OS 22.04). Embedding, retrieval testing, and vector database operations make zero outbound requests to external APIs or cloud infrastructure. Environment variables (TRANSFORMERS_OFFLINE=1, HF_DATASETS_OFFLINE=1) are set explicitly to enforce this at the system level.
  • Client documents are processed on air-gapped, encrypted local hardware and are never transmitted to third-party APIs.
  • Where performance is equal to a commercial LLM, Small Language Models (SLMs) are being actively incorporated into Funnel Amp’s client solution development and testing workflow. SMLs are preferred to LLMs due to their significantly lower energy
  • When frontier LLM inference is required for specific business development or client-requested tasks, vendor selection is governed by verified environmental criteria.
  • Conversion of the Funnel Amp office to 100% solar power is a committed future action. When completed, all local compute — including AI inference — will be powered by behind-the-meter renewable electricity, making the energy provenance of every build fully auditable.
  1. LLM Vendor Selection With a Clear Environmental Mandate

Active

Funnel Amp’s primary AI infrastructure relies on locally-deployed Small Language Models for client solution development and testing. When frontier LLM capabilities are required (for business development tasks or specific client-requested applications), vendor selection is governed by documented environmental and ethical criteria. The current approved vendor list is subject to change and expand based on the evolving qualifications of LLM vendors.

Below is the current status of Funnel Amp’s LLM vendor list:

  • First preference: Google Cloud (Gemini). Google Cloud publishes verified per-query emissions data (0.24 Wh and 0.03g CO₂e per median Gemini text prompt, as of August 2025) representing the current industry standard for emissions transparency and the closest available approximation to real-time renewable energy matching. This aligns with the standard Funnel Amp recommends to clients under this element.
  • Secondary preference: Anthropic Claude. Anthropic’s published commitments (including its Responsible Scaling Policy, Constitutional AI framework, and documented opposition to surveillance and weaponization applications) reflect values alignment with Funnel Amp’s mandate. Anthropic has not yet published Scope 1/2/3 emissions data meeting the verification standard applied to Google Cloud. This gap is acknowledged as an open accountability item, and Funnel Amp will reassess this position as Anthropic’s environmental reporting matures.

Below is the standard by which vendors are excluded from the approved vendor list:

  • Vendors without documented environmental commitments or verified emissions data are not used.
  • Vendors with documented records of infrastructure siting decisions that disproportionately burden low-income or minority communities are excluded on environmental justice grounds, consistent with Funnel Amp’s sector commitments.
  1. Documented Environmental Protection Mandates From Contractors

Active

When working with external contractors, Funnel Amp treats documented environmental protection measures, social protection measures, and carbon reduction commitments as a selection qualifier. We ask each vendor if they have a MAAS or a similar mandate in place. Priority is given to vendors who meet this requirement, while alternative contractors and vendors are selected if no qualifying match is available.

Funnel Amp’s external vendor selection process:

  • Priority is given to vendors, consultants, and developers who have a documented environmental protection mandate in place. Ideally these contractors have demonstrated proof of how they fulfill the requirements of their mandate (i.e. using renewable energy, prioritizing working with eco-responsible vendors, incorporating human-in-the-loop practices in their automation workflows, etc.)
  • Other requirements related to the quality of the vendor’s work, capability, price of services, client feedback, and availability must also be met for the vendor to be considered a viable candidate.
  • If the vendor meets all qualifications other than having an environmental protection mandate, priority will be shifted to the vendor best qualified to fulfill the work.
  1. Alignment With Applicable Regulations and Reporting Frameworks

In Progress

Funnel Amp is not itself subject to the primary regulatory frameworks governing its clients, including IRENA guidelines, CSRD, ISSB disclosure standards, or SEC ESG reporting requirements. However, these frameworks are the operational context within which all Funnel Amp client work is performed, and alignment with them is built into every engagement. As a provider in this sector, Funnel Amp engages in continuing education to ensure it remains properly informed of current and evolving regulations and frameworks.

How this element of Funnel Amp’s MAAS is applied in client engagements:

  • Each knowledge base is scoped with explicit reference to the regulatory frameworks applicable to that client’s sector and jurisdiction. For Green Energy clients, this includes IRENA guidelines and applicable grid safety and operations standards. For Sustainable Finance clients, this includes CSRD, ISSB, and SEC disclosure frameworks.
  • Metadata schemas are designed to support regulatory versioning, enabling clients to distinguish between current and superseded regulatory guidance within their knowledge base.
  • Knowledge base governance recommendations include triggers for review when applicable regulatory frameworks are updated, ensuring that AI agents do not retrieve outdated compliance guidance.
Funnel Amp - Knowledge Architecture for Responsible AI