AI case / 02 · Context assembly, fact vs inference

Preparing account context before a commercial visit

A fictional agentic case: a visit briefing a salesperson can trust—because it shows where every sentence came from.

Fictional scenario

Independently created. Contains no employer or client implementation detail, internal names or figures.

01The outcome

A one-page briefing assembled from authoritative sources with their freshness shown, every statement cited, inferences labelled, conflicts surfaced—and nothing written to any record without the salesperson.

The reality

Salespeople prepare visits by opening six screens—or not at all—and an AI summary is proposed that may invent what it cannot find.

Architecture question

Which sources are authoritative, how fresh must they be—and how does the briefing separate fact from inference?

Key decision

Assemble context deterministically from owned sources first; let the model only summarize, compare and suggest—labelled and cited—and give it no write tools.

ContextA field salesperson visits several customers a week. The information exists—account data, open opportunities, orders, invoices, performance, open actions, notes from the last visit—but in different systems, with different freshness. A generative summary is attractive and risky: it reads well even when a source is missing or stale.

Systems and partiesCRMERPData platformCalendar

02The reality · current state

What happens today—or in the pilot.

  • Visit preparation depends on each salesperson’s habits
  • Performance, orders and open actions live in different places
  • A pilot summary mentioned an order that had been cancelled
  • Nobody could tell which statements were facts and which were guesses
  • Summaries were pasted into notes and became “facts” later
  • Data conflicts between CRM and ERP were silently averaged

03Agent workflow

Each step with its actor, its tool and its authority.

  1. 01
    Upcoming visit

    A visit on the calendar for a known account triggers preparation the day before.

    SystemRead
  2. 02
    Context assembly

    Account, opportunities, orders, invoices, performance, open actions and last visit notes—each from its owner.

    SystemReadgetAccountContext
  3. 03
    Signal selection

    Only published signals (segment, coverage, changes since last visit) with their version.

    SystemReadgetSignals
  4. 04
    AI summary

    Summarizes, highlights changes and conflicts; every sentence cites a source.

    AgentRecommend
  5. 05
    Suggested focus

    Three questions to ask and open issues to close—labelled as suggestions.

    AgentRecommend
  6. 06
    Salesperson review

    The salesperson reads, dismisses or keeps suggestions; notes are written by the person.

    HumanHuman approval

04Key dimension · Context assembly, fact vs inference

Sources with standing—and a briefing that labels every line

Each source has a standing, a freshness and a use. The briefing below shows how statements are typed and cited.

Sources: standing, freshness and use
SourceStandingFreshnessUsed for
CRM account and contactsAuthoritativeLiveWho, where, relationship owner
Open opportunitiesAuthoritativeLivePipeline and next steps
ERP orders and invoicesAuthoritativeDailyWhat was bought and invoiced
Performance segmentDerivedMonthly, versionedOn Track, At Risk… with definition version
Last visit notesAuthoritativeAs writtenCommitments made last time
Email threadsExcluded—Not read: consent and relevance not established
Briefing excerpt · every line typed and cited
  • FactOrders are 18 % below the same period last year.ERP orders, refreshed this morning
  • ChangeThe account moved from On Track to Underperforming this month.Segment v3, published on the 3rd
  • InferenceThe drop is concentrated in one product line.Derived from order lines; not confirmed by the customer
  • QuestionAsk whether volume moved to a second supplier.Suggested focus—dismiss if not relevant

The pilot’s cancelled order came from a cached export. Freshness is now shown next to every source, and stale sources are left out rather than summarized.

05Authority

What the agent may do—and what it may not.

Actions and the authority the agent has for each
ActionAuthorityWhy
Read account, pipeline, orders and signalsReadWithin the salesperson’s visibility
Summarize and compareRecommendOutput is a briefing, not a record
Suggest questions and focusRecommendLabelled suggestions the salesperson may dismiss
Write visit notesForbiddenNotes are a person’s record of what happened
Update opportunities or tasksForbiddenNo write tools in a preparation task

06Ambiguity policy

When the agent is unsure.

  • Source older than its freshness ruleStop

    Left out of the briefing, with a note that it is stale

  • CRM and ERP disagreeContinue

    Both values shown as a conflict—never averaged

  • Statement cannot be citedStop

    Removed from the briefing

  • No activity since last visitContinue

    Stated as a fact, not filled with guesses

07Evaluation

Scored per dimension—never one accuracy number.

Evaluation dimensions: question, measure and target
DimensionQuestionMeasureTarget
EvidenceIs every statement cited?Statements with a valid source100 % — release gate
FaithfulnessDoes each statement match its source?Reviewer-verified statements≥ 98 %
LabellingAre inferences labelled as inferences?Correct statement type≥ 95 %
UsefulnessDid the salesperson use it?Suggestions kept or acted onTracked, no target yet

08The trade-offs

Credible options, judged against these premises.

Rejected

Let the model search every system freely

Exploration

Cost: Unknown sources, stale data, no citations
Situational

A static report, no AI

Stable, simple accounts

Cost: Changes and conflicts still need reading between screens
Selected

Deterministic assembly, AI summary with citations and labels

Visits that depend on recent change

Cost: Freshness rules and a citation format to maintain

09The second layer

Questions that change the design.

Sources

  1. Which sources are authoritative?
  2. What data freshness is required?
  3. Which sources are deliberately excluded?

Truthfulness

  1. Should recommendations distinguish fact from inference?
  2. How are citations and evidence shown?
  3. What happens when data conflicts?

Use

  1. Who may write what after the visit?
  2. How is usefulness measured?
  3. How are wrong statements reported?

10Decisions & outputs

What the work produces.

  1. 01Context model with source standing
  2. 02Freshness rules
  3. 03Citation format
  4. 04Fact / inference taxonomy
  5. 05Briefing template
  6. 06Evaluation set