Frequently asked questions.

AIBES operates as your AI department. We embed with your team to advise, build, and operate production-grade AI workflows tailored to your business, and we stay accountable for the result, not just the implementation. We are not a software vendor, and not a consultancy that hands you a deck and walks away. AIBES advises and ingrains with your strategy, engineers the AI workflows, operates them in production, and reports ROI through your control plane, AIRDEX.

It is a real role, and the load-bearing human role in our model. AI Business Engineers (AIBEs) sit at the intersection of business strategy, AI engineering, and operations. They learn the business first, then engineer AI around it. Three words capture our discipline: Advise. Build. Operate.

Mid-market companies, roughly $25M to $500M in revenue, in industries where AI has to operate inside real business processes. Our focus sectors are real estate, energy, healthcare, financial services, manufacturing, and technology. Our typical buyers are CEOs, CIOs, CFOs, COOs, and operating leaders who need AI to deliver measurable business outcomes, not pilots, not science projects.

Consultancies build and leave. Cloud vendors hand you tools and say good luck. DIY platforms work for demos and break at production. AIBES advises, builds, and operates, and stays. Our work is code-first and portable, governed end to end, and priced for production-grade ROI rather than pilot purgatory. The simplest comparison: think managed IT services, but for AI.

Four to eight weeks from kickoff to a live, governed workflow in production. Not a deck. Not a roadmap. A working system you can measure.

AIRDEX is the AI control plane every AIBES engagement runs on. This is your AI Command Center.

Real-time dashboards, approval workflows, full audit trail, AI cost intelligence, and ROI measurement, all in one place. Three layers. One AI Command Center.

• Retrieval (Knowledge): grounds AI in your authorized company data so it stops hallucinating and starts answering with your facts.
• Orchestration (Reasoning): manages multi-step logic and routes high-stakes decisions to a human reviewer when judgment matters.
• Execution (Action): connects insights to real-world systems like Salesforce, SAP, Snowflake, ServiceNow, and Teams.

Wrapped around all three layers is enterprise governance that provides policy-as-code guardrails, full-stack observability, and AI cost intelligence.

No. AIRDEX is cloud-agnostic, model-agnostic, and vendor-neutral by design. We route across providers based on cost, latency, accuracy, and compliance. You keep the freedom to change any of them later without rebuilding.

Your data stays yours, governed by your access policies. You own the workflow code we build. Deployment options range from AIBES-managed cloud to your dedicated AWS tenant. Your choice, based on your regulatory and IT requirements.

Writes only run through paths you have approved. A person approves the action in the AIRDEX interface (human-in-the-loop), or it runs through a company-sanctioned automated route you have exposed, either an API or stored procedure. Both are client approved. Every write is logged, attributable, and reversible. There is no ungoverned write path.

• Stored (persistent): the read-only replica, the retrieval index, and governed workflow memory and skills live only in the governed zone of your chosen deployment. Nowhere else.
• Processed (transient): a small, permission-checked slice of context plus the model's response, in memory only, in-flight, through AIRDEX and a zero-retention model endpoint.
• Logged (audit): who asked what, when, which records were touched, and which actions were approved, all stays within your environment.

No. We use retrieval-augmented generation, not fine-tuning. Your data stays in your governed store and is never baked into model weights. Each model call is stateless. The model sees only what it is given for that single request, and forgets it. New records are picked up on index refresh; no retraining required, no public training, ever.

Yes, deliberately. Useful agents remember prior steps, approved decisions, learned preferences, and context, because that is what makes them helpful over time. That memory is held in your single-tenant governed zone, encrypted and isolated to you, never pooled or shared. Your data, your retention, your delete button. It is never used to train a public model.

Skills are the instructions and procedures each agent follows. They are written in plain markdown files. Human-readable. Versioned. Every change is tracked: who changed what, when, and why. Every agent action is logged with full lineage from question back to source records. Changing what the AI does means changing a versioned file. It is transparent and traceable, not a mystery upgrade. It is your IP, and is exportable on request.

Yes. Single-tenant throughout. Your data, your retrieval index, and your governed memory live in a tenant isolated to you alone. It is never pooled with another client's, never used to benchmark, never used to serve other clients.

Two models, and in both cases your data sits in a single governed location isolated to you:

• AIBES-managed (recommended): a dedicated, single-tenant AIRDEX environment we provision and operate end to end. Encrypted, isolated, and deletable on request.
• Your own AWS tenant: AIRDEX runs inside your AWS account and the model is called through your enterprise endpoint. No client data is on AIBES infrastructure.

AIBES operates AIRDEX in both models, under least-privilege access you grant and can revoke.

Traditional software is deterministic: the same input gives the same output, you test it once, and it stays put. AI is not. Models drift, providers update and retire models, token costs scale with usage, and quality has to be measured continuously. Behavior changes through versioned skills and prompts, not just code deploys. Someone has to run it every day: watching quality, managing tokens, adopting new models, tuning skills, handling approvals, and keeping it secure. AIBES operates AIRDEX as a managed service, so you get the outcome without staffing an AI operations team.

• Advise: we start with your business outcomes, not the technology. Strategy, opportunity mapping, ROI modeling, and a clear-eyed roadmap are built around what your team actually does today.
• Build: we design and deliver production-grade AI workflows with the integrations, governance, and human-in-the-loop controls needed to work in the real business.
• Operate: we do not ship and disappear. AIBES operates the workflows alongside your team, measuring cost, accuracy, and adoption, and continuously tuning against the ROI promised.

We are transparent. A monthly retainer covers ongoing advisory and prioritization, with a portion applied as build credits against future projects. Each workflow is fixed-price, scoped before any development begins. Managed operations are a per-workflow subscription. The result is predictable spend that is typically a fraction of one in-house AI hire, with ROI tied directly to the workflows you turn on. Detailed pricing is shared in a scoping conversation.

AI 360 is our discovery sprint that typically takes a couple of weeks. We meet with leadership and operating teams, map your processes and data flows, and deliver a stack-ranked report of 5-10 AI use cases prioritized by ROI. One principle governs the work: every solution we recommend must deliver measurable ROI within 12 months, or we do not recommend it.

No multi-year lock-in. The retainer is month-to-month. Builds are fixed-price per workflow. Operate is a per-workflow subscription. You can pause, expand, or hand any workflow off to your own team.

Governance is defense in depth, built in at every layer. It is not bolted on. The controls include:

• Encryption: AES-256 at rest, TLS 1.2+ in transit, customer-managed keys.
• Least-privilege access: role-based access with row- and column-level policy, so users and agents see only what they should.
• Network isolation: private networking and endpoints; no public exposure of the data.
• No model training: enterprise model endpoints with zero-retention and no-training terms.
• Memory and skills governance: workflow memory and skill files are scoped, versioned, inspectable, and deletable.
• Audit logging: every query, retrieval, and agent action is logged and attributable.
• Human oversight: agents recommend, people approve. Approval gates on any action with material business impact, including write-backs.

AIRDEX is designed to align with SOC 2 and ISO 27001 control families, and we support HIPAA, GDPR, and CCPA control requirements depending on the deployment. The phrase we use: run AI like you run finance.

Policy-as-code guardrails enforce your rules at runtime. No AI action runs without clearing them. Human-in-the-loop approval routes high-stakes decisions to your team before execution. And the audit trail answers the question every CIO and CFO eventually asks: what did AI do, who approved it, and what did it cost?

AIRDEX includes full-stack AI FinOps: token tracking by model, workflow, and tenant; budget caps with model fallback; departmental chargeback; and dynamic routing that typically delivers a 30-40% cost reduction versus single-model defaults. AI costs are the new cloud bill. AIRDEX gives you the controls to manage them like one.

Book a discovery call. We will walk through the business outcomes you are after, scope an AI 360 if it makes sense, and have your first production workflow live in four to eight weeks.

Direct contacts: todd@aibes.ai and joe@aibes.ai.