Your content cannot answer your own questions
A twenty-question retrieval test settles this in days. Teams that skip it discover it during user acceptance testing, with the launch date already announced.
Agentforce · Audit & Readiness Assessment
Find out if agents will work before you buy them.
An Agentforce readiness assessment is a fixed-scope diagnostic of the four things that decide whether AI agents will work in your business: the processes you would give them, the data and knowledge behind those processes, the state of your Salesforce org, and your security posture. It ends in a costed shortlist and a recommendation. Sometimes that recommendation is not yet.
12+
500+
250+
40+
Trusted by 500+ organizations — including several who asked us to assess Agentforce and were told to fix something else first.
Assessment Scope
- Salesforce Partner
- HubSpot Partner
- Use Case Screening
- Retrieval Testing
- Org Health
- Automation Debt
- Access Posture
- Consumption Modelling
The expensive things you find out far too late
Every one of these is cheap to discover in week two and expensive to discover in month four. An assessment is a fraction of the cost of the project it might stop.
The process you picked is not actually a process
It is six people using judgement and calling it a workflow. An agent cannot follow rules that were never written, and writing them is the real project.
Automation debt makes every action unpredictable
Overlapping Flows on the objects the agent touches mean the same action produces different outcomes. The agent will be blamed for it.
The data platform cost was never in the business case
Grounding an agent in your own content depends on data platform capability. Discovering that line item after approval is a difficult conversation with finance.
Security will not approve what you are planning
Better to learn that in week two than to build for three months and meet the objection during the go-live review.
The volume does not justify the effort
A process that runs forty times a month will not repay a three-month build. Counting first is unglamorous and occasionally saves an entire programme.
What It Is
A diagnostic that is allowed to say no
An Agentforce readiness assessment is a fixed-scope, fixed-price diagnostic that answers one question: will AI agents work here, and where? It examines four things. Process fit: which work is high-volume, rule-shaped and documented enough for an agent to own. Data readiness: whether the knowledge and records behind that work can actually answer the questions, tested rather than assumed. Technical readiness: whether the org, its automation and its integrations can support an agent without being rebuilt first. And security posture: whether what you intend is something your risk function will approve.
The distinguishing feature is that this engagement can conclude that you should not proceed — or that you should fix something else first. That is not a hedge; it is the point. An assessment whose only possible output is a recommendation to buy the assessor's implementation service is a sales process with a deliverable attached.
Salesforce provides the platform and its published capabilities. Twopir Consulting provides the evidence and the judgement: a screened shortlist with rejections explained, a real retrieval test against your own content, an inspection of the org and the automation in scope, an access and controls review, a consumption model built on your actual volumes, and a written recommendation with the reasoning visible. You get a decision you can take to a board, including the option of not spending the money this year.
We run this most often for growing and mid-market companies with a decade of accumulated CRM history, where the honest answer is frequently “two of your six candidate use cases are viable, one needs six weeks of content work first, and three should never be agents.” That is a more useful outcome than an enthusiastic yes.
What we examine, how, and what you receive
Each row produces a specific artefact. If a proposal for an assessment cannot fill in a table like this, it is selling a workshop.
| Area | How we examine it | What you receive |
|---|---|---|
| Use case suitability | Request-volume analysis and transcript sampling against volume, repeatability and value. | A ranked shortlist, with rejected candidates and the reason each was rejected. |
| Knowledge readiness | A real retrieval test: index current content, run your own questions, review every passage returned. | Coverage scoring per question, plus gap, contradiction and retirement lists. |
| Data readiness | Field-level quality on the records the shortlisted agents would read; duplicate and resolution check. | A data remediation scope, limited to what the agent actually touches. |
| Org & automation health | Inventory of Flows, rules and triggers on the objects in scope; conflict and order-of-execution review. | An automation debt list with a rationalisation estimate. |
| Security & access posture | Review of the sharing model, existing permissions and what a future agent user would need. | A control gap list and an early read from your risk owner. |
| Cost & business case | Consumption modelled on your real volumes and conversation shapes, alongside delivery effort. | A cost range, a payback view, and the assumptions written down. |
| Recommendation | All of the above, weighed together with your capacity and timing. | A written go, no-go or fix-this-first, with reasoning you can challenge. |
Six investigations, run on evidence
None of these is a workshop where we ask you how things are. Each one inspects the system itself, because what a team believes about its own data is rarely what a test shows.
Use Case Screening
We count and classify what actually arrives, rather than starting from a wishlist. The output includes the candidates we reject, which is usually the more useful half of the list.
- Request volume and intent distribution
- Repeatability and documentation check
- Value and effort estimate per candidate
- Risk classification for each
- Explicit rejections with reasons
Knowledge Retrieval Test
The single most decision-changing part of the assessment. We index your current content, run your own questions against it, and show you passage by passage what an agent would have answered with.
- Test question set from real requests
- Index built over current content as-is
- Retrieved-passage review per question
- Coverage, contradiction and staleness scoring
- Data & knowledge integration →
Data Quality & Resolution Check
Scoped deliberately to the fields and objects a shortlisted agent would read. An agent programme is not a mandate for a two-year data transformation, and we will not let it become one.
- Population and accuracy on in-scope fields
- Duplicate and identity fragmentation check
- Whether one customer view exists or must be built
- Remediation effort, scoped to the agent
- What can safely be deferred
Org & Automation Review
What fires on the objects in scope, in what order, and whether any of it conflicts. Automation debt is the most common reason an agent's actions behave inconsistently in production.
- Flow, rule and trigger inventory per object
- Order-of-execution and conflict analysis
- Integration and technical debt review
- Rationalisation estimate before any build
- Automation services →
Security & Access Posture
An early read from whoever owns risk, before anything is built. The cheapest possible time to discover that your intended design will not be approved is before the design exists.
- Sharing model and permission review
- What an agent user would need per use case
- Data classification and masking implications
- Approval and evidence expectations
- Governance & security →
Consumption Model & Business Case
Cost modelled on your actual volumes and likely conversation shapes, with the assumptions stated so you can argue with them — including the data platform line that surprises most teams.
- Volume and conversation-shape modelling
- Platform and data capability cost lines
- Delivery and remediation effort estimate
- Payback view with stated assumptions
- Sensitivity if volumes are half or double
The systems we actually look inside
Read access is enough for almost all of it. We inspect rather than interview, because the gap between how a system is described and how it behaves is where the findings live.
Case & conversation history
The ground truth for what customers actually ask and how often. Volume and intent distribution come from here, not from a workshop, which is why the shortlist survives contact with reality.
Knowledge base & documents
Indexed as-is for the retrieval test. We are not judging whether the content is well written for humans — we are measuring whether it returns the right passage for a real question.
CRM records in scope
Field population, accuracy and duplication on the specific objects a shortlisted agent would read. Scoped narrowly on purpose, so the finding is actionable rather than overwhelming.
Flows, rules & triggers
Everything that fires on the objects in scope, mapped with its execution order. This is where we find the automation debt that would have made an agent's actions look unreliable.
Permissions & sharing model
How access is currently structured, and what an agent user would need on top. Orgs with heavily customised sharing often need design work here before any agent can be scoped safely.
Integrations & external systems
Where the data an agent needs lives outside Salesforce, and whether it is reachable at conversation speed. A four-second ERP call changes the design and belongs in the estimate.
Existing AI & chat deployments
What is already live, what it handles and where it fails. Sometimes the right recommendation is to fix the chatbot you have rather than replace it, and the existing data says so.
Your team's capacity
The input nobody assesses and everybody needs. A viable use case with no available content owner and no admin capacity is not viable this quarter, and the recommendation should say so.
Two to four weeks, then a straight answer
Light on your team's time — a kickoff, a couple of interviews, read access, and a readout. The work is ours; the decision is yours.
Frame & Access
Agree the candidate list and what a good outcome would look like, then get read access to the org, the knowledge base and request history.
Screen the Candidates
Volume, repeatability, value and risk for each, with rejections written down and explained rather than quietly dropped.
Test the Content
Index what exists, run the question set, review every retrieved passage. This is the step that most often changes what the client thought the project was.
Inspect the Org
Automation, data quality on in-scope fields, integrations and access posture — with the security owner involved rather than informed later.
Cost It & Recommend
Consumption modelled on your volumes, delivery and remediation estimated, and a written recommendation — go, no-go, or fix this first — presented to whoever holds the budget.
Assessments that changed the plan
Audit and readiness work is one of our most-requested engagements across Salesforce, HubSpot and AI. These are clients who bought the diagnostic before the build.
★★★★★
The AI readiness engagement gave us a clear roadmap to operationalize AI across our processes. The team built intelligent ‘Next Best Action’ capabilities using scoring, engagement, and fit models, which significantly improved how we prioritize and interact with prospects. Their understanding of both CRM and AI-driven decisioning made a real difference in aligning our systems with business outcomes.
Rubesh J. Consultant · AI readiness engagement & roadmap AI Readiness
Professional Services Firm — AI Document Automation
Automating email attachment processing and Salesforce data routing.
40% Reduction in manual document processing time
2× Admin throughput without added headcount
100% Automated classification & CRM routing
★★★★★
We engaged Twopir Consulting to conduct a Salesforce audit, and their structured, insight-driven approach exceeded our expectations. Their team quickly understood our complex processes, identified critical gaps, and provided clear, actionable recommendations. The audit improved our data accuracy, streamlined workflows, and aligned perfectly with our digital transformation goals.
Kelly Hale Managing Director · Salesforce audit & recommendations Audit
Professional Events Organization — CRM Integration
Salesforce–MeetMax integration for a corporate networking organization.
100% Elimination of manual data sync between platforms
360° Unified client view across events & CRM
0 Manual reconciliation tasks post-integration
An assessment that can cost us the build
The value of a diagnostic is entirely in its willingness to return an inconvenient answer. If it cannot, it is a proposal with a longer lead time.
No-go is a real outcome
We have told clients their content was not ready, their volumes did not justify it, and their process needed defining first. The fee is the same either way, and so is the quality of the work.
We test rather than interview
The retrieval test is the difference between an opinion about your content and a measurement of it. It routinely surprises the people who own that content.
Twelve years of looking inside CRM orgs
We have audited hundreds of Salesforce and HubSpot orgs. Recognising which findings will actually stop an agent programme, and which are cosmetic, is pattern recognition rather than a checklist.
Costs stated with assumptions attached
A number without its assumptions cannot be challenged. We show the volumes, the conversation shapes and the data platform line, so your finance team can argue with the model rather than the total.
The deliverable is yours regardless
You can take the shortlist, the retrieval results and the roadmap to another partner or run it in-house. An assessment that is only useful if you also buy the build was never a diagnostic.
Before you commission one
Four tests. Is there work that is high-volume, repeatable and already documented? Can your content correctly answer the questions that work involves — tested, not assumed? Is the org healthy enough that an agent's actions will behave consistently? And will your security function approve an agent reaching that data? Any one of those failing does not rule Agentforce out, but it changes the sequence and the cost. The assessment exists because most teams can guess at two of the four and genuinely cannot answer the other two.
Two to four weeks, at a fixed price agreed before we start, scaled to how many candidate use cases are in scope and how large the org is. The demand on your team is light: a kickoff session, two or three subject-matter interviews, read access to the org and knowledge base, and a readout. We quote it as a standalone engagement with no obligation to proceed, precisely so the recommendation cannot be read as a sales step.
Yes, when that is what the evidence says. The more common outcome is partial: some candidates are viable now, some need content or process work first, and some should not be agents at all. Occasionally the honest recommendation is that the real problem is the CRM rather than the absence of AI, and that fixing it would deliver more than any agent. We would rather say that than deliver a project that gets switched off in six months with our name on it.
No, and it is usually better if you have not. The whole point is to inform the licensing decision rather than justify one already made. Most of the assessment — request analysis, content testing, org inspection, access review — examines your existing systems and does not require the product. If you have already licensed it, the assessment still pays for itself by aiming the first build at the use case most likely to succeed rather than the one that was most appealing in the demo.
An org audit looks at everything and tells you what is not ideal. This looks narrowly at what a specific set of candidate agents would touch, and tells you whether those agents would work. It adds things a general audit does not include — a live retrieval test against your content, consumption modelling, and an agent-specific access review — and deliberately ignores org issues that have no bearing on the shortlist. Narrower scope, sharper answer, and a much shorter remediation list.
Then the assessment becomes a mid-flight review, and it is still worth running — usually as a shorter engagement focused on why progress has stalled. The two findings we most often return are that the chosen use case was not suitable, or that the grounding was never tested and the agent is retrieving badly. Both are recoverable. What is not recoverable is continuing to build for another two months on the assumption that more configuration will fix a content problem.
Next Step
Before the business case, get the evidence
Bring the use cases you are considering. We will test whether your content can answer them, inspect the org that would have to support them, model what they would cost, and give you a recommendation you can take to a board — including no.
Salesforce, CRM & AI delivery for growing and mid-market companies · Contact the team
