The short answer
A business process assessment establishes how well a specific process performs, why it performs that way, and what should change, by scoring seven dimensions (outcome, flow, ownership, systems, exceptions, controls and knowledge) with evidence for each. Horizon runs the evidence-gathering step with AI interviews across every role in the process; at AFAP SURA, that produced 63 prioritized insights across five systems in under seven days.
Performance data such as cycle time, volume and error rate is usually available. Cause and disposition require evidence about how the work actually runs, which sits with the people who do it. That is the part this framework is built to capture.
Key takeaways
- An assessment answers three questions: how the process performs, why, and what should change.
- Seven dimensions cover it: outcome, flow, ownership, systems, exceptions, controls and knowledge.
- Every score needs named evidence; where there is none, the score is recorded as unknown.
- An assessment is complete when its coverage is checked and process owners have validated it.
- The output is an as-is map, bottlenecks, ranked opportunities and inputs for formal procedures.
- Horizon gathers the evidence system data cannot supply, from every role in the process, in days instead of weeks of manual interviews.
What is a business process assessment?
A business process assessment is a structured evaluation of one process with defined boundaries. It scores how the process performs on seven dimensions, records the evidence behind each score, and names a disposition for each weak result: measure, discover, assign ownership, integrate, review controls or document.
It differs from a process audit. An audit checks whether a process followed a defined standard. An assessment asks whether the process is the right one and what should change.
When should you run one?
Three moments make an assessment worth the effort:
- Before automation. Automating a step that should not exist makes it faster and harder to remove. See process automation readiness for the conditions a process should meet first.
- Before a redesign. A redesign built on the documented process inherits every gap between documentation and practice.
- Before an ERP or system migration. A migration needs an accurate picture of the current state, including the exceptions the new system will have to handle.
An assessment also works on a smaller scope: one process in one area, ahead of a local improvement, is a good first run.
Before you start: define the boundary
Set explicit start and end points, and name the decision the assessment supports. "Vendor onboarding" can mean supplier request through first payment, or only the compliance review. A boundary drawn too wide produces vague findings; drawn too narrow, the upstream cause sits outside the frame.
The decision sets the depth. An assessment supporting a system replacement needs exception distributions and variant catalogues. One supporting a quarterly improvement cycle can work at a coarser level.
What are the seven dimensions of a process assessment?
| Dimension | Question it answers | Evidence required | If weak, the disposition is |
|---|---|---|---|
| Outcome | Does the process deliver what it exists to deliver? | The outcome metric, its current value and its variance | Measure before changing anything |
| Flow | Is the sequence known, including where work waits? | The sequence as described by the people who run it; touch time vs wait time | Discovery with the people performing the work |
| Ownership | Is responsibility clear at each step and decision? | Who actually decides at each decision point | Assign decision rights |
| Systems | Do systems support the process, or do people compensate? | Every artifact kept outside systems of record, and its hours | Integrate or consolidate |
| Exceptions | Is the variation known and quantified? | Share of volume on the standard path; distribution of the rest | Targeted discovery with exception handlers |
| Controls | Does each control still reduce a real risk? | Why each check exists and whether the reason still holds | Control review; remove what no longer applies |
| Knowledge | Can the process run without specific individuals? | Whose absence would slow the process, and whether documentation covers it | Generate documentation from practice |
Score each dimension from 1 to 5 using the conditions below. The lowest score governs the next action; an average hides the constraint.
1. Outcome
| Score | Condition |
|---|---|
| 1 | No agreed outcome measure exists |
| 3 | Outcome measured, performance inconsistent or unexplained |
| 5 | Outcome measured, performance stable and understood |
If weak, measure first. Every other dimension is harder to act on without a baseline.
2. Flow
| Score | Condition |
|---|---|
| 1 | The documented flow does not match practice, or no documentation exists |
| 3 | Main flow known, variants and wait states unclear |
| 5 | Flow known including variants, with touch time and wait time separated |
The touch and wait split is the most useful number this dimension produces, because it shows whether the constraint is capacity or queue structure.
3. Ownership
| Score | Condition |
|---|---|
| 1 | Ownership unclear; cases wait for someone to claim them |
| 3 | Steps owned, cross-functional decisions ambiguous |
| 5 | Every step and decision has a named owner who acts on it |
Assigning decision rights often reduces cycle time on its own, since a large share of wait time is cases waiting for an unclear owner.
4. Systems
| Score | Condition |
|---|---|
| 1 | Critical state held outside systems, in spreadsheets or personal trackers |
| 3 | Systems hold state, with manual bridging between them |
| 5 | Systems hold state and exchange it without manual intervention |
This dimension produces the clearest recoverable hours in most assessments.
5. Exceptions
| Score | Condition |
|---|---|
| 1 | Exception rate unknown |
| 3 | Exception types known, frequencies estimated |
| 5 | Exception types and frequencies known, with handling defined for each |
Exceptions are the most predictive dimension and the one most often skipped, because the number is uncomfortable and nobody is asked for it.
6. Controls
| Score | Condition |
|---|---|
| 1 | Controls exist with no recorded rationale |
| 3 | Rationale known, currency not verified |
| 5 | Each control traced to a current requirement with a named owner |
Assessments routinely find controls added after incidents resolved years earlier, kept because nobody had the authority to remove them.
7. Knowledge
| Score | Condition |
|---|---|
| 1 | Critical knowledge exists only with a few people |
| 3 | Documentation exists and is partly out of date |
| 5 | A capable new person could run the process from documentation |
If weak, generate documentation from how the work is actually done. Documentation written from the intended process reproduces the gap.
How do you know the assessment is complete?
An assessment can score all seven dimensions and still describe a fraction of the real process. Three checks close that gap.
1. Check coverage by role and by path. List every role that touches the process, including approvers, downstream teams and the people who handle exceptions, and confirm each one contributed evidence. Then compare the volume the assessment describes with the total: if the paths you documented account for 80 to 90% of cases, the remaining share is either documented as known exceptions or flagged as a gap.
2. Validate with process owners and leaders. Walk the process owner through the as-is map and the scores before any disposition is assigned. Ask where the map is wrong, which variants are missing, and which findings they would challenge. A finding that survives that conversation can be defended in a steering committee.
3. Flag the gaps explicitly. Where evidence is missing, record the score as unknown and say what would close it: a role not yet interviewed, a queue not yet counted, a control without a documented rationale. An unknown is an actionable finding; an estimate presented as a score gets cited later by someone who does not know it was a guess.
What outputs should you expect?
A complete assessment delivers four outputs a team can use directly:
| Output | What it contains | Who uses it |
|---|---|---|
| As-is process map | The sequence as it actually runs, with variants by team or region, exception paths and wait states | Process owners, redesign teams |
| Bottlenecks and rework | Where work waits, loops back or gets corrected, with the hours involved | Operations leaders |
| Ranked opportunities | One disposition per weak dimension, ordered by impact and effort, each with an owner | Transformation office, sponsors |
| Inputs for formal procedures | Documented steps, controls and their rationale, ready to feed SOPs and compliance documentation | Compliance, quality, risk teams |
The fourth output matters most in regulated industries. A procedure written from evidence of how the work runs, with each control traced to its reason, holds up better in an audit than one written from the intended process.
Manual assessment vs workshops vs AI interviews: which should you use?
| Method | Time to results | Coverage | Cost driver | What it misses |
|---|---|---|---|---|
| Manual assessment (document review and one-to-one interviews) | Weeks | A small sample of roles | Analyst or consultant hours per interview | Roles outside the sample; changes after the snapshot |
| Workshops | Days to weeks | The people in the room | Facilitation time and participants' hours | Exception handlers and frontline staff not invited; senior voices shape the consensus |
| AI interviews | Days | Every role in scope, in parallel | Platform and the time of the people answering | Exact transaction volumes and timings, which system data measures better |
The methods combine. AI interviews gather evidence for flow, exceptions, ownership, controls and knowledge across every role; system data confirms volumes and timings for the outcome and systems dimensions. For the full range of discovery methods, see the business process discovery guide, and for how the interviews work, AI employee interviews at scale.
Turning the result into a plan
The output of the assessment is a disposition per weak dimension, in sequence.
Measurement comes first when the outcome dimension is weak, since nothing downstream can be verified without it. Discovery comes next when flow or exceptions are weak, because the other dispositions depend on knowing what the process does. Integration, control review, decision rights and documentation follow, and can often run in parallel.
Each disposition needs an owner, an expected effect and a review date. That is what turns an assessment into a plan.
Where Horizon fits
Horizon is an AI-powered continuous discovery platform. It produces the evidence that dimensions 2 through 7 require, which is the part of a process assessment that system data cannot supply.
Discovery Cycles run AI-led interviews across every role that operates the process, adapting to each one and following up on why a step exists. The Insights Dashboard ranks findings by impact and effort, with traceability to the input behind each one, which is what makes the evidence requirement achievable. The Process Library structures the result into documentation generated from practice, which addresses dimension 7 directly and feeds formal procedures, and its process maps export to BPMN. The Initiatives Dashboard turns dispositions into business cases with owners.
What sets Horizon apart from a manual assessment is that coverage and evidence stop being a trade-off. Manual interviews take weeks and reach a small part of the team; workshops hear from whoever can attend. Horizon interviews every role that operates the process in parallel, and every finding it produces links to the input behind it, which is what the evidence rule in this framework asks for.
AFAP SURA, a pension fund administrator in Uruguay with 600 employees and part of the regional SURA group, ran this kind of assessment on two processes: ANR, covering international transfers for members living abroad, and the BPC workflows. Manual interviews had taken weeks and reached a small part of the team. Information was spread across five systems, and the team had to request status updates continuously to know where each case stood, a dimension 4 failure. An error in an international transfer could produce a financial loss that is hard to recover, a dimension 6 condition.
In under seven days, Horizon delivered 63 prioritized insights covering bottlenecks, repetitive tasks and operational risk points, with a view integrating all five systems and evidence-based recommendations to improve traceability, reduce risk and speed up processing. The manual approach had been estimated at three weeks, and the pilot returned 186% ROI.
The findings map onto the dimensions: bottlenecks are a flow finding, repetitive tasks a systems finding, and operational risk points a controls finding. Separating them turned the assessment into a set of dispositions with owners.
That is one engagement under specific conditions.
Assess one process with evidence from everyone who runs it, in one week. Run Horizon's one-week challenge: AI interviews across the process, an as-is map and a ranked list of what to fix first. Start the one-week challenge
Business process assessment template
Copy this table and complete it for one process.
| Field | Entry |
|---|---|
| Process name | |
| Boundary: starts at | |
| Boundary: ends at | |
| Decision this assessment supports | |
| Roles interviewed / roles in scope | |
| Share of volume covered by documented paths | |
| Outcome metric and current value | |
| Score: Outcome | |
| Touch time / wait time split | |
| Score: Flow | |
| Decisions without a clear owner | |
| Score: Ownership | |
| Artifacts maintained outside systems | |
| Score: Systems | |
| Share of volume on the standard path | |
| Score: Exceptions | |
| Controls without verified rationale | |
| Score: Controls | |
| Knowledge held by individuals | |
| Score: Knowledge | |
| Gaps flagged as unknown | |
| Validated with process owner (name, date) | |
| Lowest dimension | |
| First disposition and owner | |
| Review date |
FAQ
How long does a business process assessment take?
A manual assessment of one process usually takes weeks, most of it spent scheduling and running interviews. With AI interviews running in parallel, the evidence-gathering step takes days: AFAP SURA received 63 prioritized insights on two processes in under seven days.
Who should be interviewed in a process assessment?
Everyone who touches the process: the people who do each step, the ones who approve exceptions, the downstream teams that receive the output, and the process owner. Exception handlers matter most, since the exceptions dimension depends on them.
Can we start with one area or a smaller project?
Yes. An assessment works best on one process with clear boundaries, and one area is a good first scope. A small, well-evidenced assessment ahead of a local redesign is often the fastest way to show value before extending the approach.
Can we use existing manuals and procedures as input?
Yes, as a starting point. Existing manuals show the intended process; the assessment compares them with how the work actually runs. The gaps between the two are findings in their own right, and they show where documentation needs to be regenerated from practice.
What is the difference between a process assessment and a process audit?
An audit checks whether a process followed a defined standard and reports conformance or deviation. An assessment evaluates performance and cause to decide what should change, including whether the standard itself is the right one.
Should you average the dimension scores?
No. Averaging hides the constraint. A process scoring well on six dimensions and poorly on systems will not improve until the systems dimension is addressed. Use the lowest score to set the next action.
Every weak dimension leaves with an owner
A process assessment is worth running when each weak dimension ends with a disposition, an owner and a date. Score with evidence, check the coverage, validate with the people who own the process, and the result is a plan the organization can act on.
Horizon makes the evidence step fast: every role interviewed in days, and every score backed by what the people doing the work said.
See it. Fix it. Own it.