The first lesson introduced a connected workflow. The second followed a response through AI analysis and human review. This lesson answers the next question: what does the workflow need to know to interpret your evidence correctly?

For Sopact Sense, context is the information and guidance needed to understand a record, connect it to relevant evidence, and support a decision.

A document can contain part of that context. So can an earlier response, a program schedule, a definition, or an access rule. The useful step is deciding which context applies to the question being asked.

The same observation can mean different things

The training team wants to know who needs additional support before the cohort finishes. Maria’s latest record shows two missed sessions.

That observation is not yet a conclusion. Were both sessions held? Was Maria enrolled at the time? Does the program require attendance at both? Has a coach already followed up? Who is allowed to read the support note?

If one session was canceled, the record should not be interpreted as two participant absences. If both were required and missed, an agreed rule might call for a coach review. The evidence supports a different response in each situation.

Context makes those distinctions available before the team accepts an AI finding.

Five categories of context

Use these five categories to organize what your workflow needs. You do not need to supply everything your organization knows. Start with what could change the answer, its permitted use, or the next action.

1. Who or what are we following?

Define what the record represents and which relationships matter.

It may be a participant, member, applicant, grantee, supplier, site, or another organization. It may also be a program or cohort when individual identification is unnecessary.

For the training team, one person can enroll more than once. The person’s identity needs to continue, while each enrollment has its own cohort, dates, attendance, and results. Mixing those levels could attach this year’s attendance to last year’s program.

For a membership network, the relevant relationships might connect a member to a chapter, event, and certification. For a supplier team, they might connect a supplier to several delivery sites.

Write down the record you are following, how it is identified, and the relationships needed for the question.

2. What is their situation and history?

Identify the circumstances and earlier evidence that could change the interpretation.

This includes a person’s starting point, services received, previous responses, current program stage, and earlier follow-up. It can also include delivery conditions such as a site closure, staffing change, or canceled session.

For Maria, two missed sessions in week six need to be read with the schedule, enrollment dates, previous attendance, and any relevant support already offered. The same attendance pattern may require a different response early in a program than near the end.

External information can be useful when it is relevant. For example, transport disruption could help explain a pattern across one location. Keep the source and date, and treat it as background unless the evidence connects it to the particular case.

History should help explain the current record. It should not silently overwrite it.

3. How should we interpret the evidence?

Define the terms, measures, and criteria that guide the answer.

This category includes organizational priorities, data definitions, scoring rubrics, qualitative coding guidance, and relevant comparisons. It tells the workflow what counts and what does not.

For the training team, “two missed sessions” needs a definition: two consecutive required sessions that were actually held while the learner was enrolled. A canceled session does not count. A missing attendance entry is unknown until checked.

A support rule might say that this pattern triggers coach review. It should not instruct the AI to conclude that the learner has disengaged. That would go beyond the attendance evidence.

Written answers need similar care. “I feel more confident” is evidence of perceived confidence. It is not proof of employment or demonstrated skill. A coding instruction should preserve that distinction, including relevant language and local expressions.

Start with the definitions most likely to change a decision. Make their source, owner, and effective date clear.

4. What rules govern its use?

Specify who may use the information, for which purpose, and what evidence is sufficient.

Include permissions, consent, anonymity, source ownership, and the handling of missing or conflicting information. Distinguish a participant’s self-report, a verified record, and an AI interpretation.

For the training team, an assigned coach may need Maria’s identifiable support history. A funder may receive only an approved cohort summary. An anonymous experience survey should retain useful program context without being linked back to Maria.

If a rating says “no improvement” but the written answer mentions a new job, preserve both pieces of evidence and flag the conflict for review. Do not silently choose the convenient answer.

Writing a rule does not enforce it. The rule must be implemented through supported permissions or procedures and tested in the configured system. The next lesson covers how to assign that responsibility.

5. What decision or output should this support?

Name the person who needs the answer, the decision they face, and the detail they need.

The training coach needs a weekly review list: the relevant participant, dated evidence, missing information, and a question or action to consider. The program manager may need a pattern across the cohort. A funder may need a reporting-period summary with coverage and limitations.

Those outputs can use related evidence, but they have different purposes and access requirements.

A useful request is specific: “Show active enrollments needing coach review this week under the current attendance rule. Include the supporting records and unresolved gaps.”

That gives the workflow a clearer job than “Tell us who is at risk.” It also gives the reviewer a way to judge whether the answer is useful.

Context must apply at the right level

A shared definition, a cohort schedule, a participant’s history, and a report instruction should not be treated as interchangeable background.

Shared rules do not remove the need for explicit exceptions. If two applicable instructions conflict, identify the conflict and ask the responsible owner to resolve it.

Scope also matters for anonymity. An anonymous answer may carry cohort and program context while deliberately carrying no participant link.

Put the five categories together

For the training team’s question, the minimum useful context might be:

With that context, a proposed answer could say: “Maria’s enrollment needs review because two consecutive required sessions were recorded as missed. Both sessions were held. The available record does not show the reason. The assigned coach should check in.”

If one session was canceled, the same rule should produce a different finding. That is a useful test of whether the context is actually being applied.

What this means in Sopact Sense

These five categories organize the information and rules a client needs. They are not five separate product features or a guarantee that a workbook or document automatically configures the platform.

Some context is carried by the records and their history. Some is supplied through analysis instructions and reference material. Permissions must be configured through the supported controls. Your team should confirm how each important requirement is represented and test that it affects the answer as intended.

Keep the context current when a program, definition, policy, or source changes. Record which version applies and recheck the affected findings. More background is not always better; the goal is relevant, current, permitted context.

Before moving to governance

Choose one question from your workflow. Write a short answer under each of the five categories, naming the source and scope where they matter.

Then test one case whose interpretation changes with context: a canceled session, a returning participant, a changed rule, or anonymous feedback. Check whether the proposed answer uses the right evidence and respects the boundaries.

You can use the working evidence plan to record those decisions. Begin with the context needed for your first question.

Next: Govern the work. You will decide who owns the context, who may change it, and how the team reviews the results.