Decision guide

Choose who owns the failed call.

Fifteen questions, published weights and four hard gates. Unknown answers reduce confidence instead of adding points, and no score can override a missing fallback or an unsafe scope. Improving routing first is always on the table.

Published weights

Example data

tool: Bounded task 25 · Supported integration 25 · Standard policies 15 · Limited customisation 15 · Operations owner ready 20
build: Engineering capacity 25 · On-call capacity 25 · Custom orchestration need 20 · Data and integration access 15 · Budget and time 15
expert: Complex integration 25 · Limited delivery capacity 20 · Managed operations preference 20 · Funded scope 20 · Procurement owner 15

Each factor scores 0, 0.5 or 1. Scores within ten points are reported as a conditional comparison.

Calculated result

Change any input below and calculate to make it yours.

Tool score: 77.5/100
Build score: 7.5/100
Expert score: 57.5/100
Evidence confidence: Medium

Gates

Safe scope: Clear. No unsafe tasks declared.
Human fallback: Clear. Named owner covers every hour the agent runs.
Verifiable outcome: Clear. Outcomes can be checked in a system record.

Findings

i
Scores: Tool 77.5, Build 7.5, Expert 57.5
Unknown points — Tool 0, Build 0, Expert 20.
i
Keep the current process / improve routing first
Always a legitimate outcome. Compare it in the cost model before committing to any route.

Score contributions

15 rows

Route	Factor	Weight	Answer value	Points
Tool	Bounded task	25	1	25
Tool	Supported integration	25	0.5	12.5
Tool	Standard policies	15	1	15
Tool	Limited customisation	15	1	15
Tool	Operations owner ready	20	0.5	10
Build	Engineering capacity	25	0	0
Build	On-call capacity	25	0	0
Build	Custom orchestration need	20	0	0
Build	Data and integration access	15	0.5	7.5
Build	Budget and time	15	0	0
Expert	Complex integration	25	0.5	12.5
Expert	Limited delivery capacity	20	1	20

Responsibilities that stay with you

3 rows

Route	Retained responsibilities
Tool	Testing on your calls; Caller policies; Fallback staffing; Monitoring; Contract review
Build	Telephony; Speech and model components; Reliability; Releases; Monitoring; On-call support
Expert	Scope approval; Access decisions; Acceptance testing; Supplier accountability

30-day plan

4 rows

Step
Week 1: run the readiness scorecard and model completed-task cost.
Week 2: shortlist two configurable services with documented integrations.
Week 3: run identical trial cases on both.
Week 4: decide on verified completion and cost per completed task.

Assumptions behind this result

Published weights: Tool — bounded task 25, supported integration 25, standard policies 15, limited customisation 15, operations owner 20. Build — engineering 25, on-call 25, custom orchestration 20, data access 15, budget/time 15. Expert — complex integration 25, limited delivery capacity 20, managed operations 20, funded scope 20, procurement owner 15.
Each factor scores 0, 0.5 or 1. Unknown answers add uncertainty, never points.
Scores within ten points produce a conditional comparison.
Hard gates override every score. No route supplies legal clearance.

Next diagnostics for this route

Readiness plus economics.

Readiness scorecard Completed-task cost Providers suited to tool

Inputs

Label each value: Provided, Extracted, Assumed or Unknown. Unknown values are excluded and reported, never treated as favourable. Use aggregate or redacted data only — no recordings, patient records or payment details.

Workload

Workflow

Call direction

Eligible calls per month

Sensitive data on the call

Implementation budget (planning)

Urgency

Tool readiness

Is the task bounded: a short list of allowed outcomes and clear exclusions?

Does an off-the-shelf service document an integration with your scheduling or job system?

Are your caller policies standard enough to configure without custom logic?

Can you live with the configuration limits of a packaged tool?

Is someone ready to own testing, monitoring and fallback staffing day to day?

Build readiness

Do you have engineers with telephony or real-time audio experience available?

Can your team run on-call support for a production calling system?

Does the workflow need orchestration no packaged tool supports?

Do you control API access to the systems the agent must read and write?

Is there budget and time for a multi-month build before results?

Expert readiness

Is the integration complex: several systems, legacy access or unusual rules?

Is your internal delivery capacity limited for this project?

Would you rather a specialist operated and monitored the agent?

Is there funded scope for a specialist implementation?

Is a named person ready to own supplier procurement and accountability?

Gates

Does the scope include emergency triage, clinical or financial advice?

Is a named person covering handoffs whenever the agent runs?

Can you verify the outcome in a system (booking, job, ticket)?

Outbound calling authorisation review

Runs in your browser. Nothing is uploaded.