Learning path · Level 9 · Tutorial 2

Comparing AI and Non-AI Solution Options

Compare AI and non-AI delivery options for a shortlisted task and recommend an approach using explicit requirements, evidence, access constraints, pricing structures, and switching risks.

  • Level: Intermediate
  • Duration: 18 min
  • 8 steps
Start tutorial Back to the path
Comparing AI and Non-AI Solution Options

What you will go through

  1. Set the Requirements Before Comparing Tools Define Cedar Office Supplies' technology-neutral comparison brief before evaluating any AI or non-AI option. 2 min
  2. Compare Four Ways to Deliver the Same Outcome Compare four delivery approaches for the same policy-grounded reply task while preserving the common requirements and staff approval boundary. 2 min
  3. Separate Evidence from Claims Classify observed behavior, documented commitments, unverified claims, and missing evidence while keeping each conclusion within its recorded scope. 3 min
  4. Check Team Access and Integration Fit Identify team-access and integration dependencies that could prevent an otherwise capable option from meeting Cedar Office Supplies’ common brief. 2 min
  5. Read Pricing Without False Comparisons Identify billing units, known subtotals, pricing dependencies, and unanswered pricing questions before comparing solution options. 2 min
  6. Look Beyond Export Buttons Assess switching risk by checking which workflow artifacts can actually move, be reused, or must be rebuilt. 2 min
  7. Build the Decision Matrix Construct an evidence-linked matrix that exposes requirement failures, unresolved questions, and meaningful tradeoffs. 3 min
  8. Recommend an Approach and Explain the Tradeoffs Synthesize Cedar Office Supplies’ completed decision matrix into an evidence-based recommendation that preserves unresolved limits and distinguishes further consideration from deployment approval. 3 min

What you will learn

  • Build a decision matrix comparing an existing assistant, a no-code solution, a specialized vendor, and a non-AI alternative against the same task requirements.
  • Distinguish verified capabilities and documented commitments from unsupported claims or missing evidence.
  • Identify access restrictions, pricing dependencies, and portability barriers that could undermine an otherwise attractive option.
  • Recommend an approach and explain its tradeoffs, including when a non-AI solution is sufficient.

Before you start

  • Prioritizing AI Use Cases for a Team
  • Configuring a Reusable Assistant Workspace
  • Building a No-Code Workflow with a Conditional Branch
  • Creating an AI Privacy and Security Checklist

Step 1 of 8

Set the Requirements Before Comparing Tools

Define Cedar Office Supplies' technology-neutral comparison brief before evaluating any AI or non-AI option.

Define the outcome before the tools

Start with the shared task

Before comparing tools, write one brief that every option must serve. Cedar Office Supplies needs draft replies to routine delivery and returns questions for 4 staff handling 400 inquiries/month.

Permitted input: the 5 approved policy documents, within the existing data-handling restrictions.

Required output: a draft reply that staff can review, approve, and copy as plain text before sending.

This is a task definition, not a tool choice. Each later option must be tested against this same outcome and boundary.

A common requirements map

A diagram showing five policy documents feeding a draft reply, followed by a staff approval checkpoint and plain-text output for four staff members. Inbox transfer and tone adjustment appear as side options.

The central path is required for every candidate. The side branches are conveniences, not a new definition of the task.

Tip

One brief, one approval boundary

Keep the required outcome and the human review boundary fixed. A tool may prepare a draft, but a staff member approves it before it is sent. Comparing different tools against different outcomes would make the comparison unfair.

Write a technology-neutral brief

Separate needs from preferences

A must-have requirement is an observable need that every acceptable approach must meet. An optional benefit would improve the process but is not required for the task to work.

Write requirements in technology-neutral language. Describe what people must be able to do and what the process must produce—not a preferred product, model, connector, or feature. A named feature can be considered later, but it must not redefine the task.

Cedar's requirement categories

Use this common brief before looking at any product features.

The same categories apply to every option considered later.
NeedCategoryTechnology-neutral wording
Individual access for 4 staffMust-have requirementEach staff member can use the process.
5 approved policy documentsMust-have requirementDrafts use the approved policies.
Unclear questionsMust-have requirementUnclear questions are referred to staff.
Approval before sendingMust-have requirementStaff approval is required before sending.
Plain-text transferMust-have requirementApproved drafts can be copied as plain text.
Inbox transferOptional benefitThe process transfers an inquiry automatically.
Tone customizationOptional benefitStaff can adjust the draft's tone.

Example

Observable need versus feature preference

Good requirement: “Staff can approve a policy-based draft before sending.”

Feature preference: “The process must use a named AI feature.”

The first can be checked across different approaches. The second prematurely picks a technology and may exclude a sufficient non-AI process.

Practice: protect the common brief

Classify the comparison brief

Match each statement from Cedar's brief to its category.

Tap an item and then its matching pair.

Step 2 of 8

Compare Four Ways to Deliver the Same Outcome

Compare four delivery approaches for the same policy-grounded reply task while preserving the common requirements and staff approval boundary.

Keep the outcome constant

One task, one approval boundary

The outcome is fixed: Cedar Office Supplies needs draft replies to routine delivery and returns questions using 5 approved policy documents. All 4 staff need individual access; unclear questions go to staff; staff approval is required before sending; and approved drafts must be copyable as plain text.

Compare a complete delivery approach, not an isolated feature. Every option needs a route from permitted policy material to draft preparation, human review, and usable output transfer. The tool or process may change, but the must-have requirement and approval boundary do not.

Four routes to the same approved reply

A four-lane diagram showing an inquiry, approved policies, a different preparation method, staff review, and plain-text output in every lane.

A — Existing assistant, B — No-code solution, C — Specialized vendor, and D — Non-AI alternative can differ in preparation method while retaining the same policy source, human review, and plain-text transfer boundary.

Map the complete route

Compare whole workflows

The following are four possible operating designs. They are not evidence that a particular product already provides every step. Later, you will test each design against evidence and constraints.

Four complete approaches

Each row includes the work around the drafting component—not just the component itself.

Conceptual routes for the same Cedar Office Supplies task
ApproachComplete route
A — Existing assistantStaff accesses permitted policy material, provides relevant context to the assistant, reviews the draft, and copies approved text.
B — No-code solutionA no-code workflow uses permitted policy inputs and an AI-enabled step to prepare a draft; staff reviews it and transfers approved text.
C — Specialized vendorAn AI-enabled support product uses an authorized policy source to prepare a draft; staff reviews it and transfers approved text.
D — Non-AI alternativeStaff uses shared approved reply templates and deterministic rules to prepare a draft; staff reviews it and copies approved text.

Tip

Avoid category assumptions

“No-code” describes how a workflow may be assembled, and “specialized vendor” describes a product category. Neither label guarantees task capability, better results, or less human review.

Treat D as a credible delivery option

When a non-AI approach can be enough

D — Non-AI alternative is a real candidate, not a failure to modernize. It uses shared approved reply templates and deterministic rules rather than AI-generated wording.

It can be sufficient when routine inquiry patterns can be directed to an appropriate approved template without interpretation beyond the rules. It must still preserve the same boundaries: use approved policies, refer unclear questions to staff, require staff approval before sending, and allow plain-text copying.

Example

Example: templates plus rules

A fictional routine inquiry asks about returning an unopened item. A deterministic rule directs the staff member to the relevant approved return-reply template. The staff member fills in permitted case details, checks the draft against the approved policies, and approves or edits it before copying the text.

If the inquiry has an unclear exception or does not fit a rule, it is referred to staff rather than forced into a template.

Tip

Drafting is not sending

Automation can change who prepares the first draft. It does not remove the need for human review when the task requires staff approval before sending.

Check your comparison

Choose the complete approach

Which conclusion is the fairest comparison for Cedar Office Supplies?

Step 3 of 8

Separate Evidence from Claims

Classify observed behavior, documented commitments, unverified claims, and missing evidence while keeping each conclusion within its recorded scope.

Classify evidence before drawing conclusions

Evidence answers a bounded question

For Cedar Office Supplies, evidence must answer a specific question about preparing draft replies from 5 approved policy documents. Do not rank evidence labels from “best” to “worst.” Instead, ask what a source actually establishes.

  • Observed behavior shows what happened in a stated example.
  • Documented commitment records a feature, access term, or service term.
  • Unverified claim is asserted without adequate supporting evidence.
  • Missing evidence means the available packet does not answer the question.

A successful draft in one example supports a narrow task-capability observation. It does not establish reliability across varied inquiries, errors, exceptions, or staff correction needs.

Evidence has different scopes

A diagram of four evidence sources: one observed screen example, a plan document, an unsupported promotional claim, and an empty evidence folder, each linked to a different scope boundary.

Classify the source, then state the limited conclusion it supports.

Tip

Evidence record

Record enough context for another person to check the claim later: Source, Date, Plan/version, and Scope. If a detail is not recorded, write that it is not recorded; do not silently fill the gap with an assumption.

Read the packet without upgrading its claims

Cedar evidence packet

This compact packet uses the same shortlisted task and approval boundary for all options. It is selection evidence only: none of these entries proves production reliability.

Evidence packet

Each row preserves both the evidence category and its limit.

Fictional selection evidence for Cedar Office Supplies
Evidence labelSourceDatePlan/versionScopeWhat it establishes
Observed behaviorInternal task observation2026-03-01A workspace; version not recordedOne routine delivery inquiry using approved policiesA produced one draft; not consistency across inquiries
Observed behaviorInternal template walkthrough2026-03-01Existing office license; version not recordedOne routine return inquiry and one unclear-question referralD followed its templates and deterministic rules in these examples
Documented commitmentB workflow plan page2026-03-01B Business planIndividual accounts and approval-step configurationB documents these workflow features; no observed Cedar task result
Documented commitmentC commercial terms2026-03-01C workspace planStated user allowance and draft-reply allowanceC documents commercial terms; not task quality
Unverified claimC sales presentation2026-03-01C workspace planClaim of accurate policy-grounded drafts; scenarios and results absentA claim needing corroboration
Missing evidenceEvidence review log2026-03-01C workspace planVaried inquiry results, error patterns, and correction needsReliability and human review effort remain unresolved

Keep capability, reliability, and review evidence separate

Check the conclusion

A documented statement that B has an approval-step configuration proves that B has already drafted Cedar’s routine inquiries accurately.

Apply the evidence labels

Classify evidence and reject overreach

Match each fictional excerpt to the most accurate evidence category and scope conclusion.

Tap an item and then its matching pair.

Step 4 of 8

Check Team Access and Integration Fit

Identify team-access and integration dependencies that could prevent an otherwise capable option from meeting Cedar Office Supplies’ common brief.

1. Test access beyond a personal trial

Check access as a team capability

For Cedar Office Supplies, Team access means more than proving that one person can open a tool. All 4 staff need permitted individual access to the features used in the process, the shared approved resources they need, and the review route.

Check these separately:

  • individual accounts for all 4 staff;
  • shared administration and team-level controls;
  • access to the 5 approved policy documents or shared templates;
  • restrictions tied to plan, administrator approval, region, or organization policy.

Keep the candidate configuration precise. For B — No-code solution, the candidate is B Workflow Pro. A personal trial can reveal a limitation of that trial, but it does not establish the status of B Workflow Pro. Existing data-handling restrictions remain binding at every handoff; this check applies them without reopening the full assessment.

Individual access is different from shared access

A comparison of one personal account with four individual staff accounts using shared work resources under organization controls.

One personal account, shared resources, and team-level administration are different access questions. The brief requires permitted individual access for all 4 staff.

Tip

Avoid a common shortcut

A paid-only administration feature or connector is an access dependency. It may be acceptable, but record it before treating the feature as available to the team.

2. Separate required handoffs from conveniences

Inspect every handoff

Map the whole delivery path, not only the candidate tool. Cedar’s path includes the inquiry inbox, the 5 approved policy documents, the candidate approach, staff approval, and a plain-text reply.

At each handoff, ask:

  1. Can information move by a permitted route?
  2. Who can perform the step?
  3. Is a connector available for the applicable plan, organization, and region?
  4. If not, is permitted manual transfer acceptable?

Automatic inbox transfer is an optional benefit. The must-have requirement is still met when staff can use a permitted manual route, unclear questions go to staff, staff approve before sending, and the approved draft is copyable as plain text.

Required route and optional connector

A workflow from an inbox and approved policy documents to a candidate tool, staff approval, and a plain-text reply, with a dashed optional inbox connector.

The solid route can use permitted manual transfer. The dashed inbox connector may reduce steps, but it is optional and can depend on plan, organization, or regional enablement.

3. Read the plan and scope, not just the feature name

Selection snapshot for access and handoffs

Read access evidence at the stated plan/version and scope. In this selection snapshot, B Workflow Pro is the B candidate—the same configuration used in the decision matrix.

Meets means that the displayed selection evidence supports the stated requirement within its scope. It does not validate deployment or prove that every optional connector is available. This snapshot concerns access and handoffs, not detailed pricing or configuration work.

Plan-feature table

The core route allows manual transfer. The automatic inbox connector remains an optional benefit, even when its availability is unresolved.

Illustrative selection evidence for Cedar Office Supplies. B refers to B Workflow Pro unless a different configuration is explicitly named.
CriterionRequirementABCD
Team accessIndividual access for 4 staff using permitted accountsMeets — Documented commitment: A plan sheet; 2026-03-01; A Team; 4 individual accountsMeets — Documented commitment: B workflow guide; 2026-03-01; B Workflow Pro; at least 4 permitted individual member accounts and member rolesMeets — Documented commitment: C service terms; 2026-03-01; C Team; up to 4 usersMeets — Documented commitment: D office record; 2026-03-01; existing office license; 4 staff accounts
Integration fitStaff approval and plain-text output; inbox transfer optionalMeets — Observed behavior: A task check; 2026-03-01; A Team; staff hold and plain-text copyMeets — Documented commitment: B workflow guide; 2026-03-01; B Workflow Pro; reviewer step and plain-text output; connector not recordedMeets — Documented commitment: C service terms; 2026-03-01; C Team; reviewer queue and text copyMeets — Documented commitment: D office record; 2026-03-01; existing office license; staff approval and manual plain-text copy

Example

A trial result is not a plan result

Keep configurations separate

A configuration comparison note dated 2026-03-01 records Observed behavior for a different setup: a B personal trial allowed only its creator, and personal accounts could not hold the approved policies. That is a blocking access gap for that trial configuration.

It does not change the candidate status for B Workflow Pro. For B Workflow Pro, at least 4 permitted individual member accounts are a Documented commitment in the B workflow guide. Its automatic inbox connector is still Missing evidence for Cedar’s organization and region, but that connector is optional because the core route can use permitted manual transfer.

4. Identify what blocks, what needs evidence, and what is optional

Identify what blocks, what needs evidence, and what is optional

Using the plan-feature table and the handoff diagram, which conclusion correctly identifies the blocking access gap, the unresolved dependency, and the optional integration?

Step 5 of 8

Read Pricing Without False Comparisons

Identify billing units, known subtotals, pricing dependencies, and unanswered pricing questions before comparing solution options.

Read the units before the numbers

Start with the billing unit

A price is meaningful only with its billing unit. Cedar Office Supplies has 4 staff and handles 400 inquiries/month, but an inquiry is not automatically a user, a workflow run, or a draft reply.

Before comparing options, name what is billed: a seat, a workspace, a run, a draft-reply allowance, or a separate provider service. A quoted subscription is not the same as the complete operating cost.

Pricing anchors

Illustrative pricing, not current vendor offers.

Keep each billing unit separate; do not convert inquiries into runs or draft replies without evidence.
OptionQuoted pricingBilling unit or allowanceKnown limitation
A — Existing assistant$20.00/user/monthIndividual user4 users can produce a license subtotal
B — No-code solution$30.00/workspace/month + $0.02/runWorkspace and runsSeparate AI-provider fees unspecified
C — Specialized vendor$120.00/workspace/monthUp to 4 users and 500 draft repliesOverage pricing unspecified
D — Non-AI alternative$0.00/monthSoftware subscriptions under existing office licenseStaff effort is still required

Expose dependencies and unknowns

Calculate only what the evidence supports

A simple license subtotal is valid only when both the price and quantity are known. For A — Existing assistant: 4 individual users × $20.00/user/month = $80.00/month.

That is a known subscription subtotal, not a complete operating-cost estimate. For B — No-code solution, the workspace fee is known, but the number of billable runs is not. The separate AI-provider fees are also unspecified. Do not assume 400 inquiries means 400 runs.

For C, 500 draft replies are an included allowance, not evidence that Cedar will create 500 billable replies. For D, $0.00/month in added subscriptions does not mean no staff work is needed.

Pricing dependencies

A four-path diagram showing that option B depends on both workflow runs and a separate AI provider, while the other options use different billing structures.

A dependency diagram reveals charges and unknowns that a single advertised price can hide.

Tip

Keep unknowns visible

Record unanswered questions rather than filling gaps with assumptions. For B, ask what event creates a billable run, whether retries create additional runs, and what the separate AI provider charges. For C, ask what happens above the included draft-reply allowance. These questions clarify pricing structure; they do not produce a complete operating-cost estimate.

Check a like-for-like conclusion

Apply the pricing evidence

Using only the illustrative pricing, which statement is correct?

Step 6 of 8

Look Beyond Export Buttons

Assess switching risk by checking which workflow artifacts can actually move, be reused, or must be rebuilt.

Portability is more than an export

From export to portability

An export button answers only one question

A claim such as “export available” does not establish low switching risk. First ask: Which artifact can leave, in what format, and can the next approach use it?

For Cedar Office Supplies, approved reply text may be easy to copy as plain text. That is useful, but the working approach also depends on the 5 approved policy documents, reusable instructions or rules, workflow configuration, and possibly run history.

Portability is about preserving useful work—not merely producing a file. Greater dependence can be acceptable, but its benefit, limitation, and evidence should be explicit before commitment.

Artifact-to-destination map

A diagram contrasts artifacts that can move directly with workflow configuration and history that may need rebuilding or verification.

Check each artifact separately: direct transfer, rebuilding effort, or unresolved portability.

Check the artifacts, format, and destination

What needs to move

Use an artifact-by-artifact check. A file is portable only when the destination can meaningfully use it; a vendor-specific export may be readable but still require manual reconstruction.

Record a focused evidence request with its Source, Date, Plan/version, and Scope. If no answer or sample is available, label the point Missing evidence rather than assuming either success or failure.

Portability check

Illustrative portability questions for Cedar Office Supplies

Artifacts have different reuse paths and different switching effort.
ArtifactWhat to verifyPossible result
Approved draft repliesCan staff copy them as plain text?May move directly
5 approved policy documentsCan the next approach receive the current files?May move directly
Reusable instructions or rulesCan they be exported in a usable form?May need rewriting
Workflow configurationAre steps and mappings visible outside the tool?May need rebuilding
Run historyCan it be exported, and is it usable elsewhere?May be unavailable or limited
Connector setupIs it proprietary, plan-limited, or reusable?May create switching effort

Example

Turn a claim into a testable request

Example: a narrow claim, not a portability finding

A product statement says, “Approved draft replies can be exported.” This supports only the export of approved reply text within the stated plan and scope. It does not show whether reusable instructions, workflow configuration, connector settings, or run history can move to another approach.

A focused request could be: “Please provide a sample export of the workflow configuration, identify its file format, and show whether its connector mappings can be reused outside this product.”

Until that request is answered for the relevant plan/version, those artifacts remain Missing evidence.

Practice: avoid the export-button shortcut

Check your reasoning

A tool exports approved draft replies as a CSV file. No information is provided about its reusable instructions, connector mappings, or workflow configuration.

Does the CSV export alone prove that the complete reply-preparation workflow is portable?

Portability check

Does this export alone prove that the complete workflow is portable?

Apply the switching-risk check

Make a portability judgment

For Cedar Office Supplies, explain why exporting approved draft replies as plain text does not by itself prove that a reply-preparation workflow is portable. Then name one specific artifact or format that should be verified before commitment.

Write at least 80 characters (0/80).

Step 7 of 8

Build the Decision Matrix

Construct an evidence-linked matrix that exposes requirement failures, unresolved questions, and meaningful tradeoffs.

Start with a like-for-like frame

Use one comparison frame

A decision matrix compares each approach against the same brief rather than turning unlike facts into one unexplained score. Use the fixed columns Criterion | Requirement | A | B | C | D and the same seven criteria for every option.

Each cell has a status and a short evidence note. Meets means the displayed selection evidence supports the stated requirement within its scope; it does not mean approved for deployment. Does not meet requires evidence of a confirmed failure. Unresolved means the evidence is incomplete, narrow, or does not establish the full requirement.

Keep the common must-have requirement set fixed: individual access for all 4 staff; use of the 5 approved policy documents; referral of unclear questions to staff; staff approval before sending; and approved drafts copyable as plain text. Automatic inbox transfer and tone customization are optional benefits.

Status meanings

A status is an evidence-linked judgment, not a product rating.

Use the most cautious status that the stated evidence supports.
StatusMeaningWhat the note retains
MeetsThe requirement is supported within the evidence scopeSource and relevant scope limit
Does not meetA stated requirement is confirmed absent or failsFailure evidence and source
UnresolvedThe packet cannot support a full conclusionGap, narrow scope, or unsupported claim

Link each judgment to evidence

A seven-row, four-option decision matrix with evidence cards connected to individual cells.

A matrix cell preserves both its judgment and the evidence limitation behind it.

Read the evidence packet

Use the packet consistently

Read the selection evidence before assigning a status. A limited observation can support a narrow behavior, but it cannot establish reliability across Cedar Office Supplies’ 400 inquiries/month. A documented feature can support an access or review commitment, but it does not substitute for an observed task result.

The candidate configurations are A Team, B Workflow Pro, C Team, and D under the existing office license. Prices below are Illustrative pricing, not current vendor offers. An inquiry, a run, and a draft reply are different billing units.

Compact selection-evidence packet

All entries are fictional selection evidence for Cedar Office Supplies.

Each record identifies its evidence label, Source, Date, Plan/version, and Scope.
Evidence labelSource and DatePlan/versionScope
Observed behaviorA task check; 2026-03-01A TeamOne routine delivery draft used approved policies; staff held and copied plain text; unclear-question referral not tested
Documented commitmentA plan sheet; 2026-03-01A Team4 individual accounts; $20.00/user/month
Documented commitmentB workflow guide; 2026-03-01B Workflow ProPolicy-input configuration, reviewer step, plain-text output, and at least 4 permitted individual member accounts
Documented commitmentB pricing sheet; 2026-03-01B Workflow Pro$30.00/workspace/month + $0.02/run; separate AI-provider fees unspecified
Documented commitmentC service terms; 2026-03-01C TeamUp to 4 users, reviewer queue, text copy, $120.00/workspace/month, and 500 draft replies
Unverified claimC sales presentation; 2026-03-01C TeamAccurate policy-grounded delivery and returns drafts claimed; no scenarios or results supplied
Observed behaviorD template walkthrough; 2026-03-01Existing office licenseOne routine return used an approved-policy template; one unclear question was referred; staff approved and copied text
Documented commitmentD office record; 2026-03-01Existing office license4 staff accounts, shared templates, manual plain-text copy, and $0.00/month added subscriptions
Missing evidenceSelection review log; 2026-03-01All four candidate configurationsVaried-case results, error patterns, and correction needs
Missing evidenceSelection review log; 2026-03-01All four candidate configurationsUsable export formats or direct reuse evidence for instructions, rules, configuration, connectors, or history
Missing evidenceSelection review log; 2026-03-01C TeamOverage pricing after 500 draft replies is not recorded

Tip

Keep gaps visible

Do not fill gaps with assumptions. A requirement with two parts needs evidence for both. A’s task check supports one policy-grounded routine draft, but it does not test referral of an unclear question. If the matrix requirement includes both behaviors, that cell is Unresolved.

A documented commitment that B Workflow Pro provides at least 4 permitted individual member accounts supports its Team access requirement. However, that access evidence does not establish B’s task capability. An optional benefit cannot cancel either a confirmed must-have failure or an Unresolved must-have requirement.

Model short, traceable notes

Keep unlike judgments separate

Build the matrix in short sections. Keep performance, price structure, and portability separate: a stated subscription price is not evidence of task capability, and an export claim is not evidence of low switching risk.

For the pricing row, use this requirement: Material provider, usage, and overage charges are known. This does not require a total operating-cost estimate. It does require that material charge dependencies are known. Therefore, unspecified separate AI-provider fees or unspecified overage pricing leave the relevant cell Unresolved.

No candidate configuration in this packet has a confirmed must-have failure, so none of these model cells uses Does not meet. That does not make every option eligible: an Unresolved must-have requirement still needs clarification.

Worked matrix sections

These rows model concise, traceable notes. The evidence packet supplies the full Source, Date, Plan/version, and Scope for each note.

Selection-evidence judgments only. Meets does not validate deployment.
CriterionRequirementABCD
Task capabilityPolicy-grounded routine draft and unclear-question referralUnresolved — A task check; referral not testedUnresolved — B guide; no Cedar task result or referral evidenceUnresolved — C claim; task performance unverifiedMeets — D walkthrough; routine template and unclear referral
Human reviewStaff approval before sendingMeets — A task check; staff holdMeets — B guide; reviewer stepMeets — C terms; reviewer queueMeets — D walkthrough; staff approval
Team accessIndividual access for all 4 staffMeets — A plan sheet; 4 accountsMeets — B guide; at least 4 permitted individual member accountsMeets — C terms; up to 4 usersMeets — D office record; 4 accounts
Pricing dependenciesMaterial provider, usage, and overage charges are knownMeets — A plan sheet; $20.00/user/monthUnresolved — B pricing sheet; AI-provider fees unspecifiedUnresolved — C terms; overage pricing unspecifiedMeets — D office record; $0.00/month added subscriptions

Apply the matrix to all four approaches

Build the complete evidence-linked matrix

Create Cedar Office Supplies’ complete seven-row decision matrix using the evidence packet. Use these exact criterion labels: Task capability, Reliability evidence, Human review, Team access, Integration fit, Pricing dependencies, and Switching risk.

For each criterion, state the requirement and give A — Existing assistant, B — No-code solution, C — Specialized vendor, and D — Non-AI alternative a status plus a short source-based note. Use Meets, Does not meet, and Unresolved precisely.

Then add two concise conclusions:

  1. Identify which option or options currently meet all common must-have requirements within the selection evidence, and identify the must-have gaps or unresolved points for the others.
  2. Identify one unresolved tradeoff for each option. Keep pricing dependencies separate from task-performance and switching-risk judgments. Do not use numeric scores or make a deployment recommendation.

Write at least 700 characters (0/700).

Step 8 of 8

Recommend an Approach and Explain the Tradeoffs

Synthesize Cedar Office Supplies’ completed decision matrix into an evidence-based recommendation that preserves unresolved limits and distinguishes further consideration from deployment approval.

Use evidence to frame the decision

Turn the matrix into a recommendation

A matrix is a decision aid, not a contest for the most automated option. Start with the same must-have requirement set for every approach: 4 staff need individual access; drafts use the 5 approved policy documents; unclear questions are referred to staff; staff approve before sending; and approved drafts are copyable as plain text.

Then state a proportionate recommendation:

  1. Name an approach for further consideration, or explain why commitment should wait.
  2. Connect its principal benefit and tradeoff to matrix evidence.
  3. Explain why alternatives are not preferred now.
  4. Name unresolved questions and the evidence needed to answer them.

An optional benefit, such as automatic inbox transfer or tone customization, cannot cancel an Unresolved must-have requirement. A recommendation for further consideration is not approval to deploy. Existing data-handling restrictions remain binding.

One decision, one approval boundary

Four delivery approaches use the same policy documents and converge at staff approval before a plain-text reply is transferred.

Compare benefits and limits against the same outcome. Evidence gaps remain visible instead of being replaced with assumptions.

Example

Read the evidence within its scope

Compact selection-evidence recap

  • A — Existing assistant: Observed behavior in the A task check, dated 2026-03-01, plan/version A Team: one routine delivery draft used approved policies; staff held and copied plain text. Referral of an unclear question was not tested.
  • B — No-code solution: Documented commitment in the B workflow guide, dated 2026-03-01, plan/version B Workflow Pro: policy-input configuration, reviewer step, plain-text output, and at least 4 permitted individual member accounts. No Cedar task result or unclear-question referral is observed.
  • C — Specialized vendor: Documented commitment in C service terms, dated 2026-03-01, plan/version C Team: up to 4 users, reviewer queue, text copy, pricing, and a 500-draft-reply allowance. Its policy-grounded task claim is an Unverified claim in a sales presentation.
  • D — Non-AI alternative: Observed behavior in the D template walkthrough, dated 2026-03-01, existing office license: one routine return used an approved-policy template, and one unclear question was referred to staff.

For all options, varied-case results, error patterns, correction needs, and usable reuse/export formats remain Missing evidence.

Use the matrix as the decision boundary

Use the completed matrix consistently

This recap preserves the completed matrix’s requirements, candidate configurations, and evidence-linked statuses. Meets means supported within the stated selection evidence; it does not establish deployment readiness. No new evidence is introduced here.

Completed decision matrix recap

Selection-evidence recap for Cedar Office Supplies.

The same criteria and requirements apply to A, B, C, and D. Prices are Illustrative pricing, not current vendor offers.
CriterionRequirementABCD
Task capabilityPolicy-grounded routine draft and unclear-question referralUnresolved — A task check; referral not testedUnresolved — B guide; no task result or referral evidenceUnresolved — C claim; task performance unverifiedMeets — D walkthrough; routine template and unclear referral
Reliability evidenceEvidence across varied cases, errors, and correction needsUnresolved — selection review log; no varied-case evidenceUnresolved — selection review log; no observed task resultUnresolved — selection review log; claim is not reliability evidenceUnresolved — selection review log; no varied-case evidence
Human reviewStaff approval before sendingMeets — A task check; staff holdMeets — B workflow guide; reviewer stepMeets — C service terms; reviewer queueMeets — D walkthrough; staff approval
Team accessIndividual access for all 4 staffMeets — A plan sheet; 4 accountsMeets — B guide; at least 4 permitted individual member accountsMeets — C terms; up to 4 usersMeets — D office record; 4 accounts
Integration fitStaff approval and plain-text output; manual inbox transfer acceptableMeets — A task check; hold and plain-text copyMeets — B guide; reviewer step and plain-text outputMeets — C service terms; reviewer queue and text copyMeets — D office record; approval and manual plain-text copy
Pricing dependenciesMaterial provider, usage, and overage charges are knownMeets — A plan sheet; $20.00/user/monthUnresolved — B pricing sheet; AI-provider fees unspecifiedUnresolved — C terms; overage pricing unspecifiedMeets — D office record; $0.00/month added subscriptions
Switching riskUsable export or direct reuse evidence for working artifactsUnresolved — selection review log; reuse evidence absentUnresolved — selection review log; configuration reuse absentUnresolved — selection review log; configuration reuse absentUnresolved — selection review log; template/rule reuse unverified

Tip

Keep pricing dependencies separate

Do not turn the pricing row into a total-cost claim. A has a known license subtotal of $80.00/month for 4 users. B has a $30.00/workspace/month fee plus $0.02/run, while separate AI-provider fees are unspecified. C has $120.00/workspace/month for up to 4 users and 500 draft replies, while overage pricing is unspecified. D adds $0.00/month in software subscriptions under the existing office license.

An inquiry, a run, and a draft reply are different billing units. The 400 inquiries/month do not establish how many runs or draft replies will be billed.

Explain tradeoffs and evidence requests

Model recommendation: consider D, do not deploy yet

A defensible current conclusion is to recommend D — Non-AI alternative for further consideration, not for deployment. D is the only option whose selection evidence currently supports the full task-capability requirement, including a routine approved-policy template and referral of an unclear question. It also meets the recorded human review, team access, integration-fit, and pricing-dependency requirements.

Its principal benefit is a straightforward process for routine delivery and returns questions using shared approved reply templates and deterministic rules. Its tradeoff is narrower handling of variation: staff may need to choose and adapt templates manually, and automatic inbox transfer and tone customization remain optional benefits.

Do not state that D has low switching risk. Its reusable template and rule formats remain Unresolved, as does its reliability across varied inquiries. The recommendation is proportionate because D is supported for this narrow task, while its important limits remain visible.

Example

Reject alternatives for evidence-based reasons

Why the other options are not preferred now

  • A — Existing assistant: Its task check supports one routine delivery draft, but the unclear-question referral was not tested. That leaves a must-have requirement Unresolved. Its known 4-user license subtotal is $80.00/month.
  • B — No-code solution: Its documented workflow features support human review, integration fit, and individual access for all 4 staff. However, there is no observed Cedar task result or evidence of unclear-question referral, and separate AI-provider fees are unspecified.
  • C — Specialized vendor: Its reviewer queue, text copy, and access allowance are documented, but its policy-grounded task-performance statement is an Unverified claim. Overage pricing after 500 draft replies is also not recorded.

A cautious alternative conclusion could be to delay commitment entirely, because reliability evidence and switching risk are Unresolved for every option. That conclusion is not anti-AI; it follows the evidence limits.

Name questions that could change the decision

Before any commitment, make focused evidence requests:

  1. Reliability question: Can D handle a broader set of routine delivery and returns questions, identify exceptions, and show staff correction needs? Evidence needed: recorded task observations with Source, Date, Plan/version, and Scope.
  2. Switching risk question: Can D’s templates and deterministic rules be exported or directly reused in another approach? Evidence needed: a sample artifact and its usable format for the existing office license.
  3. If reconsidering B or C: Does the applicable configuration support the full unclear-question referral route, and what charges apply beyond the stated pricing anchors? Evidence needed: applicable-plan documentation and task evidence, not a general sales claim.

These requests reduce uncertainty. They do not approve deployment or replace the existing data-handling restrictions.

A sound comparison sequence is: common requirements → evidence and limits → team access and integration fit → pricing dependencies → switching risk → proportionate recommendation.

Comparison sequence review

Use this sequence to compare a technology choice with a non-AI alternative fairly.
CheckDecision discipline
Common briefKeep the same must-have requirement and human review boundary for A, B, C, and D.
EvidenceUse Observed behavior, Documented commitment, Unverified claim, and Missing evidence within their stated scope.
Access and costKeep individual access and pricing dependencies visible; inquiries, runs, and draft replies are different units.
PortabilityAn export claim does not settle switching risk for instructions, rules, configuration, connectors, or history.
RecommendationA non-AI approach can be sufficient; further consideration is not approval to deploy.

Final recommendation application

Apply the full comparison

Using the matrix recap, write a short recommendation for Cedar Office Supplies in three labeled sections.

  1. Chosen approach and rationale: Choose A, B, C, D, or recommend delaying commitment. Tie your conclusion to the must-have requirements and available evidence.
  2. Tradeoffs and alternatives: State one principal tradeoff. If you choose A, B, C, or D, give a requirement- or evidence-based reason for not preferring each of the other three options now. If you recommend delaying commitment, explain why none of the four options is ready to be preferred now.
  3. Unresolved questions and required evidence: Name at least two unanswered questions and specify what evidence would answer each before commitment.

You may recommend the non-AI approach, an AI-enabled approach, or delay. Do not treat a recommendation for further consideration as approval to deploy.

Write at least 500 characters (0/500).

Tutorial complete

Congratulations! You completed: Comparing AI and Non-AI Solution Options

You can now compare AI and non-AI options using common requirements, evidence limits, access constraints, pricing dependencies, switching risk, and a proportionate recommendation.

Download the App now to have access to + 5000 free courses, exercises, certificates and lots of content without paying anything!

  • 100% free online courses from start to finish

    Thousands of online courses in video, ebooks and audiobooks.

  • More than 60 thousand free exercises

    To test your knowledge during online courses

  • Valid free Digital Certificate with QR Code

    Generated directly from your cell phone's photo gallery and sent to your email

Cursa app on the ebook screen, the video course screen and the course exercises screen, plus the course completion certificate