Answer balance, transaction, standing-order and card questions on your channels once the session is authenticated — complaints, fraud and vulnerability route to named staff, and the agent never moves money.
Take the question on the bank's own channels — app message, web chat, secure inbox or the servicing queue.
02
Authenticate the session to the bank's agreed standard before any balance or transaction is read back.
Reason
03
Read the customer's own record — balances, posted and pending items, standing orders, direct debits and card status.
04
Answer from the tariff and terms in force for the product actually held, and keep the clause the answer rests on.
05
Check the servicing state that changes the answer — arrears, a hold, a block, a live dispute or a bereavement marker.
Decide
06
Detect a complaint, a fraud disclosure, a hardship or a vulnerability signal and route it to the team that owns it.
07
Hold anything the approved answer set does not cover, or that would commit the bank to a fee, a rate or a term.
Out
08
Answer on the channel, or hand the session to a named person with the record and the reason already gathered.
09
Retain the question, the authentication, the sources read, the answer given, the routing and every correction.
→Product statement
The agent answers inside an approved set on an authenticated session. Money movement, fee decisions and complaint outcomes stay with named bank staff.
Example workflow
One servicing contact, end to end
AgentHuman
1Contact receivedApp message, web chat, secure inbox, an IVR handoff or the servicing queue
2Session authenticatedChecked to the bank's agreed standard, with step-up before anything sensitive is read back
3Record and terms readBalances, posted and pending items, standing orders, direct debits, card status and the tariff in force
4Answer drafted and checkedChecked against the approved answer set, the servicing state on the account and what the bank could be held to
No human action required
Stages 1 to 4 run before anyone at the bank opens the session — authentication, retrieval and the checks finish first. A complaint, a fraud disclosure or a vulnerability signal ends that stretch on the spot.
5DecisionSplits on whether the contact is inside the approved answer set
Inside the answer set
Answered on the channel, with its source kept.
Outside it, or a duty is engaged
Routed to the team that owns it, with the clock.
Banker, complaint handler or specialist team
Reads the question, the record the agent read and the answer it drafted, then replies, logs the complaint against its clock, or takes the case on.
Answer · Correct · Escalate
Answered — handed back▼
6Answer sent or case routedSent on the channel, or routed to a named team with the date of receipt and the reason recorded
7Outcome evaluatedAnswer accuracy against the record, routing fit, complaints logged late, repeat contacts and corrections by cohort
Corrections
What a person rewrites before the customer sees it is counted.
What should not run autonomously
Human approval stays in control
Outside the boundary — human approval required8 items
Moving money, or changing a payment instruction.
Closing, blocking or reinstating an account or card.
Waiving, refunding or applying a fee.
Resolving a complaint, or deciding one was not made.
Automation boundaryAgent acts unaided
✓Authenticate the session before reading anything back.
✓Show balances, transactions and card status.
✓Answer standing-order, direct-debit, statement and tariff questions from the record.
✓Route a complaint, a fraud disclosure or a vulnerability signal to a person.
Write actions run only inside the approval boundaries agreed during implementation. Moving money is not one of them.
Deciding a disputed transaction or a fraud claim.
Taking on a bereavement, hardship or vulnerability case.
Advising on a product, a debt or an investment.
Changing the answer set or the authentication rules.
Example output
One servicing contact, annotated
Everything the agent answers is attached to the session and the record it came from.
Servicing output · single authenticated contactIllustrative example
Customer wrote
Channel
Session
Routed to
Confidence
Fee decision
Why was I charged? This isn't right
App message, authenticated
Step-up passed
Complaints, clock started
89%
Not made by the agent
As receivedThe customer's own words and the authenticated channel they arrived on — nothing on this side is rewritten.
Evidence usedTariff in force, §4Posted card transactionArrears marker on file
Why it was routed, not answeredThe wording is dissatisfaction, so it is logged as a complaint with the date received.
ActionAnswerCorrectEscalate
What the score decidesHow closely the handler re-checks the tariff before replying — not whether the fee was fair.
Value
Where AI adds value
The same four claims, placed at the point in the workflow where each one applies.
Where the value landsValue 01 – 04
Every servicing contactApp, web chat, secure inbox or phone
03Record & terms
Answer from the account
Use the customer's own balances, transactions and instructions, and the tariff and terms for the product actually held.
01Approved path
Clear the answerable contacts
Balance, transaction, statement, standing-order and card questions are answered where the customer asked, instead of queueing behind everything else.
02Human review
Get duty cases to a person
A complaint, a fraud disclosure, a bereavement or a hardship signal reaches the team that owns it with the record and the date of receipt already recorded.
04Build an evidence trail
Retain the question, the authentication, the sources read, the answer given, the routing and the correction — on both paths.
Integrations
Typical integrations
Five system groups connect to the same agent. Which of them are in scope is decided in discovery.
Core bankingFIS · Fiserv · Jack Henry Temenos · Core servicing APIs
Customer channelsMobile app · web chat Secure inbox · telephony and IVR
Failures cluster in the contacts that carry a duty
A balance question on a single-holder account is close to a lookup — one record, one answer. What goes wrong is the complaint never phrased as one, and the disclosure made in passing. Nestack reports by slice, not only in total.
Slice performance — reported separately, not only in aggregateIllustrative example
Slice
Failure rate
Lift
Lift vs. threshold
Status
Complaints phrased as questions
5.7%
3.4×
Review
Vulnerability disclosed mid-contact
4.6%
2.8×
Review
Joint and authorised-user accounts
3.2%
1.9×
Watch
Routine balance and transaction queries
1.1%
0.7×
Normal
Bar: mishandled-contact lift vs. routine-query baseline · scale 0–4.0× · tick marks the 2.0× review threshold2 of 4 slices over threshold
Evidence-linked improvement
A missed duty shows up in the transcript, not the dashboard
What tells you a contact went wrong is usually a second contact, a complaint logged late or an ombudsman file — weeks after the session that caused it.
Improvement cycle · five stagesSwitchback — the path turns at Improve and returns at Learn
01Detect
A repeat contact, a late-logged complaint or an escalation traces back to one session.
02Diagnose
Pulled apart at the session — the gate, the record used, the wording, or the rule that should have routed it.
03Improve
The answer set, the routing rule or the wording is re-approved by compliance and version-linked.
04Verify
Replayed against the sessions that were missed and a held-out sample from the same cohort.
05Learn
The wording nobody caught is added to the routing tests, and the session is kept as a case.
Learn → DetectThe return edge. A routing rule decides who a customer reaches, so complaints and the vulnerable-customer team sign off the change before it ships.
Typical build scope
Twelve workstreams across six weeks
The build scope read against the delivery timeline. Week structure follows the six-week plan — discovery, the answer set and authentication, complaint and routing rules, evaluation, integration, then production validation and handover.
WorkstreamWeek 1Week 2Week 3Week 4Week 5Week 6
01Workflow discovery and automation-boundary definition.
02Core-system, channel and CRM assessment.
03Approved answer set and out-of-scope rules.
04Authentication and step-up rules.
05Complaint identification and clock capture.
06Tariff, terms and product-document store.
07Account, instruction and card-status reads.
08Fraud, bereavement and vulnerability routing.
09Answer-accuracy and routing evaluation suite.
10Language access, accessibility and channel testing.
11Channel and case-management integration.
12Observability, deployment and Agent Care handover.
12 workstreams · 6 weeks · bar shows the weeks a workstream is active — several run in parallelFinal scope and sequence confirmed in discovery
Engagement tiers
What each tier includes
Rows are the capabilities named in each tier's scope. Higher tiers include everything below them.
Capability✓ in scope · — not at this tierPilotOne servicing channelProductionProduction core integrationAdvancedMulti-brand / multi-channel
Introduced at Pilot
Answers from the customer's own record✓✓✓
Authentication and step-up gate✓✓✓
Complaint, fraud and vulnerability routing✓✓✓
Audit trail of answers and routing✓✓✓
Baseline evaluation✓✓✓
Introduced at Production
Wider product and tariff coverage—✓✓
Additional servicing channels—✓✓
Case-management and CRM integration—✓✓
Observability and evaluation—✓✓
Introduced at Advanced
Multi-language and accessibility coverage——✓
Multi-brand and enterprise controls——✓
Build priceFrom $5,000From $8,000Custom quote
Final build priceConfirmed after discovery based on servicing channels, core and case-management integrations, product and tariff breadth, contact volume, approval controls and deployment requirements.
Separate from buildBuild pricing is separate from recurring Agent Care, which covers managed monitoring, evaluations, incidents and verified improvements after launch.
What we need from you
What you bring, and what we build with it
Each input maps to a piece of build scope and a week in the delivery timeline.
You bringWe build with it
01The questions your servicing queue actually gets→Approved answer set and out-of-scope rulesWeek 1
02Your authentication standard and when you require step-up→Authentication and step-up rulesWeek 2
03Access to core, channel, CRM and case-management systems→Core-system, channel and CRM assessment, then integrationWeek 2
04Your complaint definition, its clocks, and the ones found late→Complaint identification, clock capture and regression casesWeek 3
05The tariff, terms and product documents in force→Tariff, terms and product-document storeWeek 3
06Where fraud, bereavement, hardship and vulnerability cases go→Specialist routing and the handover into your queuesWeek 4
07Named handlers, and the languages and channels your customers use→Language and channel testing, then supervised live sessionsWeeks 5–6
Nothing else is requiredDeployment, documentation and Agent Care handover are ours.
Delivery timeline
Four phases across six weeks
Phases are drawn over the weeks they actually occupy. Week 5 carries both the language and channel testing and the first supervised sessions.
PhaseW1W2W3W4W5W6
DiscoveryW1
BuildW2 – W3
EvaluateW4 – W5
Pilot & LaunchW5 – W6
Week focusW1The questions in scope, the ones that are not, and where each goesW2Authentication, step-up and access to core, channel and CRMW3Complaint identification, the clock, and the tariff and terms storeW4Specialist routing, evaluation suite and routing testsW5Channel integration, supervised sessions and language testingW6Customers are answered on one live channel, then handover
Reading the bandComplaint identification and the routing tests finish in week 4, before a customer reads an agent answer in week 5. The bars show that order, not a smooth ramp.
At the end of W6The agent has answered on one live channel under supervision, and routed cases reached the teams that own them, then Agent Care takes over monitoring.
DurationSix-week plan shown · typical delivery 4–6 weeks depending on scope confirmed in discovery.
Next step · Banking AI agent
Build a servicing agent around your own answer set.
Show us your servicing queue, your authentication standard and what your complaint policy counts as a complaint. We'll take a week of real contacts, mark which ones the agent may answer and which carry a duty, and show you where the clock would have started.
