Agentforce products you sell. Production agents you run.
Aquiva builds Agentforce two ways. ISVs hire us to ship agents as products: packaged for the AgentExchange, security-reviewed, and sustained through every platform release. Enterprise teams hire us to put agents into their own org: grounded in their data, governed by the trust layer, with evals that prove the agent stays on-script. More than ten Agentforce projects shipped, and an authorized member of Salesforce's Approved Forward Deployed Engineering (FDE) PDO Partner Network.
Salesforce Cloud Expert PDO · Select Partner · 10 yearson platform · 70+ verified reviews at 4.91/5
ISVs and Salesforce customers. Either way, the agent ships with evals, versions, and an audit trail.
Whether the agent is sold or run in your own org, it ships with evals that catch regressions before a customer does, prompts under version control, and an audit trail that survives a compliance review.
ISVs shipping agents other companies install
Your agent installs into many customer orgs: topics, actions, prompt templates, and Agent Templates in a managed package. Customers create an agent from your template instead of rebuilding it step by step, the security review now covers the agent, and three platform releases a year still land on your subscribers. The packaging, listing, and sustain journey lives with our PDO practice; this page is the Agentforce layer on top: planners, eval harnesses, prompt versioning, and multi-channel delivery.
Enterprise teams running agents in their own org
The agent works inside the org your people already run, so it has to honor your data model, sharing rules, and audit trail. We configure the trust layer, build eval harnesses before the agent faces a customer, and version prompts so a small tweak can't quietly regress last month's accuracy. Actions that spend money or change records get human-in-the-loop approval. Internal agents pointed at your own team live with Custom Solution Development, and org design lives with Salesforce consulting.
Named artifacts you can version, test, and operate.
Whether it lists on the AgentExchange or runs inside one org, the artifact ships versioned, tested, and observable.
Agentforce products for the AgentExchange
Customer-facing agents an ISV sells. Topics, plugins, prompt templates, and packaged Agent Templates, in 1GP or 2GP, delivered across chat, voice, Slack, and embedded experiences. Security-review prep runs with Application Security.
Production agents in your org
Topic mapping, action definition, guardrails, and persona design, inside your permissions and audit trail. Headless where the UI isn't Lightning: API-first interfaces, MCP servers, tool registration, and multi-agent orchestration.
Grounding: Data 360 and the trust layer
Every Agentforce LLM action routes through the Einstein Trust Layer; we configure masking, grounding, prompt defense, and audit, and we name what it doesn't cover. Variables and filters keep an action from running until verified data says it should. When identity resolution has merged two customers into one profile, the agent answers confidently from the wrong person's data, so the data model gets fixed before the agent ships. The deeper data-layer work lives with Data & Integrations.
Security of the agent surface
Indirect prompt injection is a live attack path: ForcedLeak hid instructions in Web-to-Lead data and an agent followed them. We run Agentforce Security Posture Assessments across agents, tools, prompts, data flows, Trusted URLs, and action scopes, with least privilege as the default. The code-level work lives with Application Security.
Buy the outcome, or run the roadmap.
Delivery follows the same shapes as everything we build. Three fit agent work best.
Outcome
A defined artifact with a date: the listing, the production agent, the security-review pass. Scope, price, and delivery are ours to hit.
AI POD
An Architect/Orchestrator, two Developers, and a Product Owner shipping in 1–5 day cycles at a fixed monthly rate. You steer the roadmap; the POD ships it.
Forward Deployed Engineering
Salesforce selected Aquiva as an authorized member of its Approved Forward Deployed Engineering (FDE) PDO Partner Network: engineers who move Agentforce from pilot to production inside customer environments. Connecting agents to disconnected systems, structuring data so retrieval works, and putting controls in place before a demo becomes an incident. From AI pilot to production →
The numbers we steer by are published on How we work: 87% of pilots reach production, and production lands in under 90 days.
Named customers, verified reviews.
Halosight
An AgentExchange listing we took through a first-pass security review, now rated 5.0/5 across verified reviews. The product evolved into an AI agent for outside sales: meeting prep, note-taking, and per-account chat on Service Cloud, Data 360, and Tableau CRM. Case study →
Conexiom
Touchless document automation on the AgentExchange: exceptions and trading-partner linking stay in Salesforce. Case study →
Agentforce patterns
The solution briefs cover agent work as repeatable patterns: patient visit management, document workflows and e-sign, multilingual support, candidate matching packaged for review, case summaries for service teams. Solution briefs →
The PDO muscle underneath
RainFocus, BetterUp, Hapi, and Kaptio hired us to build products their customers use; the same discipline ships agents. Case studies →
“Their guidance was instrumental in navigating the Salesforce Security Review process — including multiple security scan iterations, addressing findings, and preparing the necessary compliance documentation. They also provided valuable input on AppExchange publishing, pricing models, and go-to-market strategy. The project was delivered on time and met every key milestone.”
Verified project review · Across AI AppExchange launch · December 2025 · Read it on our AgentExchange listing
Common questions
Does Aquiva implement Salesforce Agentforce?
Yes, in two forms. If the agent is a product other companies install, that runs through our PDO practice with the Agentforce layer built here. If it runs in your own org, we design the topics and actions, configure the trust layer, and stand up the evals and human-in-the-loop controls it needs in production.
What's the difference between Agentforce for ISVs and for enterprise teams?
An ISV ships a managed package of topics, actions, and Agent Templates onto the AgentExchange, through security review, into many customer orgs. An enterprise team configures agents against one org's data model, sharing rules, trust layer, and evals. The artifacts overlap; the install surface is what changes.
Can Agentforce agents be packaged for the AgentExchange?
Yes. Actions, topics, prompt templates, and Agent Templates are packageable in 1GP and 2GP, so customers create an agent from your template instead of reconstructing it by hand. Security-review preparation runs with our application security practice.
How do Agentforce Flex Credits work for ISV partners?
Platform (OEM) partners can resell Flex Credits at wholesale and carry the overage liability, and only in OEM-only orgs where they are the sole seller: once the org becomes shared, with Salesforce selling into it directly, the OEM can no longer submit credit orders. ISVforce partners ship into shared orgs and can't resell credits; their customers buy them from Salesforce. One partner can hold both contract types, and the right structure is a pricing decision worth making before the listing ships.
What is the Einstein Trust Layer, and can ISVs package it?
The trust layer sits between org data and external models: masking, grounding, prompt defense, toxicity scoring, and audit. Every Agentforce LLM action routes through it. ISVs can't package it, since each customer's org provides its own; we configure what fires and name the gaps it leaves, like masking coverage and the Data 360 cost of audit and grounding.
How do you keep an agent from going off-script?
Specific topics, instructions that name the output format and the missing-data behavior, and variables and filters so actions only run when verified data supports them. Prompt changes go through eval harnesses before release, tool scopes stay least-privilege, and actions that spend money or touch customers keep a human in the loop.
How long does it take to get a live agent?
The numbers we publish: 87% of pilots reach production, and production lands in under 90 days. A POD ships in 1–5 day cycles, and a defined listing or agent can run as an Outcome with a date attached.
Let’s put an agent into production.
Whether it ships to the AgentExchange or into your own org, tell us what you're building. Most engagements start with a 30-minute call.