Strategy and architecture across the four AI tracks — agents, retrieval, automation, integrations — followed by build inside your delivery process. Not a research project. Not a chat box.
Hub work, then build inside the right track
These are the conversations agency owners describe when an AI brief is in front of them and the in-house team is two people deep on prompt experiments but hasn't yet shipped a production system.
The pitch landed. The buyer is excited. The spec is one paragraph. We come in for a feasibility readout — what's actually buildable, on what timeline, against which model, with what guardrails — before the SOW gets locked at the wrong shape.
The demo got applause. Production wants evals, fallback paths, rate limits, citation discipline, and a fixed response time. We re-architect from a demo into a system that survives a billing-day incident review.
n8n + a model + good intentions only carries a workflow so far. We come in to redesign for reliability — orchestration, retries, observability, ownership boundaries — without throwing away the working pieces.
Demand is recurring. The team can't keep saying yes on a single specialist. We stand AI Implementation up as a new service line via Capability Expansion — capability lead, pricing template, delivery bench behind it.
Most AI briefs we see are a mix of two of these tracks. The hub work decides which one is primary, which one is supporting, and which one is being asked for but isn't actually the answer.
Tool-using · Multi-step
Production agents that take actions, call tools, and run multi-step workflows — with evals, guardrails, and a fallback story when the model is wrong.
Pick this when When the brief is "AI that does things," not "AI that answers things."
Retrieval · Citations
Retrieval-augmented generation built around your client's content — with chunking strategy, reranking, citation discipline, and an honest "I don't know" path.
Pick this when When the answer must be grounded in the client's content, not the model's training set.
Internal · Reliable
Internal automations that take a manual, repeated workflow off your client's plate — designed for reliability, not for a one-off demo.
Pick this when When the win is hours saved per week, not a new client-facing surface.
HubSpot · Slack · CRMs
Wiring AI into the systems your client already runs in — HubSpot, Slack, CRM, custom internal tools — instead of building a parallel chat surface nobody opens.
Pick this when When the right answer is AI inside the existing tool, not a new tool to learn.
The hub engagement decides which track owns the build. Most briefs end up combining two of these — the architecture readout in week one names the primary and the supporting.
See the full capability benchAI work runs feasibility-first. The hub engagement decides whether the brief is buildable and which child capability owns the build, before any production code ships.
Working session with your delivery lead and the buyer-side stakeholder. We pressure-test the brief against current model capability, compliance posture, and your team's existing stack. NDA and SOW signed under Salt Technologies, Inc.
A short, written architecture readout: chosen track (or tracks), model and routing strategy, evals, guardrails, observability, and the things we are explicitly not building. Your team and your client refer back to it for the rest of the engagement.
Build runs inside Agents, RAG, Workflow Automation, or Integrations — under the same SOW, the same delivery lead, and the same pod. Weekly working review with named owners on both sides. Evals tracked from build one.
Production deployment, observability dashboards, cost monitoring, and a written runbook your team can operate without us. Post-launch support transitions cleanly into a Support & Maintenance retainer for model drift, dependency upgrades, and v1.x feature work.
AI Implementation is most often launched as a new service line via Capability Expansion. The other two shapes are the entry points when a live brief is already on the table.
Capability lead + delivery support
Standing AI Implementation up as a defensible new service line on your rate card — capability lead, pricing template, monthly retainer with quarterly review.
Pre-sales feasibility scoping
An inbound AI brief you want to win — needs senior feasibility, an architecture point of view, and a defensible estimate inside 72 hours.
AI pod inside your SOW
Active AI client work that has been sold and now needs end-to-end throughput inside your shared channel — long-arc, structured, named delivery owner.
Starting rate for a single-capability AI pod. Multi-track pods (e.g. RAG + Integrations) and pods that mix capabilities are quoted at the highest applicable rate. Final SOW is scoped against the brief.
We work inside the AI tooling your team and your client already use — no parallel platform, no "we'll just rebuild it our way" surprise.
AI Implementation is the most-requested capability on the bench right now. These three agency profiles describe where it shows up most often as the lead.
HubSpot partners shipping AI workflows, retrieval-grounded portals, and AI integrations across the HubSpot stack — where the standard tooling ends and a real AI build begins.
Boutique design and product studios shipping AI-native product surfaces for clients — where the brief asks for an interface that thinks, not a new chat tab.
Shopify partners adding AI inside merchant tools, shopper assistants, and content-grounded support flows that need real RAG and integration work behind them.
Every AI Implementation engagement runs on the same operating spine that protects long-arc retainers and Dedicated Partner Pods — contracted through Salt Technologies, Inc.
We don't email your client, join their calls, or appear in the proposal — unless you explicitly white-list a named engineer in the SOW.
We work in your GitHub, your model-provider accounts, your hosting, and your shared channel under aliases that fit your team's naming.
Mutual non-solicitation written into every MSA, with a defined window after the engagement ends. Same clause across every track.
MSA, NDA, and engagement SOW are issued by Salt — the Delaware C-Corp behind AgencyRelay.
The same operating spine sits underneath every AgencyRelay capability. Read the no-poach and confidentiality page for the contractual instruments behind these defaults.
Most agency owners deciding between an AI pod, a single specialist hire, or a freelance prompt engineer weigh the same trade-offs. These pages address them head-on.
The most common shape for adding AI Implementation to your rate card — capability lead, pricing template, quarterly review on whether the new line earns its place.
When an AI pod is cheaper, faster, and lower-risk than the next senior AI hire — and when staffing up is the right call.
When the AI brief lands inside a live pitch and you need feasibility, architecture, and a defensible estimate inside 72 hours.
Direct answers to the questions that come up on almost every AI Implementation scoping call.
Because the four AI tracks behind it solve different problems and use different toolchains. Agents take actions, RAG grounds answers in content, Workflow Automation runs reliable internal pipelines, and Integrations wire AI into existing tools. Most briefs combine two of those — the hub work figures out which combination, then build runs inside the right child capability. One page, one rate, one pod, four possible builds.
Prompt engineering is one piece of one of the four tracks. AI Implementation engagements are end-to-end systems work — feasibility, architecture, model and routing strategy, evals, guardrails, observability, integration, runbook. Prompt design is part of the work; it isn't the work.
Senior. Every AI Implementation engagement is staffed with engineers who have shipped real AI systems to production inside agencies and product companies — not a junior pool with a senior badge in the project channel. If continuity is ever at risk on a long-arc engagement, we tell you before the next sprint review.
By default the client's. We work inside their OpenAI, Anthropic, or other model-provider accounts under named seats. Where you (the agency) hold the account on behalf of the client, we operate inside yours. Either way, the account ownership and cost-attribution boundary is documented in the SOW so there's no ambiguity at month-end.
Sometimes — but it's almost never the right answer. If the brief deserves a real production system, we say so before the calendar invite goes out. Most demo-shaped engagements end up needing rebuild work for evals, guardrails, and integration; we'd rather route the brief into a Proposal Rescue Desk feasibility window first and ship the right thing once.
Yes — they're part of the architecture readout in week one. Model selection, data routing, retention, PII handling, and (where applicable) HIPAA / SOC 2 / EU data-residency posture are written into the readout. We work inside the compliance envelope your client already operates in; we don't invent a parallel one.
The starting weekly rate for a single-capability AI pod is $3,200 per week. Multi-track pods (e.g. RAG + Integrations) and multi-capability pods (AI + UI/UX, AI + Backend) are quoted at the highest applicable rate. Final SOW is scoped against the brief; the rate is the floor, not a ceiling.
Most AI systems graduate cleanly into a Support & Maintenance retainer post-launch — model drift monitoring, prompt regression on new model versions, cost optimisation, eval upkeep, and v1.x feature work inside a monthly envelope. Either the same pod or a smaller maintenance crew carries it on the same MSA.
Yes. IP ownership and assignment on delivered code, prompts, evals, and supporting artefacts is written into the MSA — the work belongs to your agency (and onwards to your client per your own client contract) on payment of the relevant invoice. The Salt Technologies templates are counsel-reviewed and shared before signing.
Agents, RAG, workflow automation, integrations, or all four. Either way, the conversation starts with the work — not with a deck.
