Engagement patterns, honestly labelled.
Quenora Consulting is a founder-led firm in its early chapter. Rather than fabricate a client roster or borrow logos we have no right to, the below describes the shape of the work — the structures, the constraints, and the kinds of outcomes these engagements are built to produce.
How the work takes shape.
Our capabilities can be applied across a wide range of operational and technical problems. The map shows that breadth; the three examples below show what it can look like in practice.
These are illustrative scenarios, not limits on the work we take on, and the figures shown are not audited client results.
Nine capabilities, and the places the work lands
Invoice processing across three ERP instances
A mid-size financial services group processing high invoice volumes across three unintegrated ERP systems, with a finance team manually reconciling between them. The engagement focused on the integration layer rather than replacing any of the three systems.
Support triage and knowledge retrieval
A manufacturer whose service desk was routing tickets by hand and losing institutional knowledge as experienced staff retired. We built a retrieval layer over two decades of resolved tickets and wired it into the existing service desk rather than replacing it.
Legacy modernisation and predictive scheduling
A logistics operator running scheduling on a system older than most of its staff. Rather than a rewrite, we built an integration and data layer that let modern predictive models reach the legacy core safely.
Why this page looks different
As a young firm, we do not yet publish named client case studies. Rather than fill that gap with vague or overstated claims, we have chosen to show the structure of the work, the problems we address, and the way we approach delivery.
As engagements are completed and approved for publication, this page will evolve to include named case studies. Where appropriate, reference conversations can also be arranged during the briefing process.
Plausible is not the same as correct
Four questions someone might reasonably ask an internal AI assistant. In each example, both answers sound fluent and plausible. The difference is that one is grounded in the available information, while the other quietly invents details that are not there.
The dashed figures highlight the fabricated values.
“Invoice INV-88214, €48,900 from Rehder GmbH. It doesn’t match the PO. Can I post it?”
Ungoverned
—▶ answering
Governed
—The honest cost: the governed answer takes roughly 2.3s longer and reads two documents to do it. That is the price, and we state it rather than hide it. Documents, figures and reference numbers here are worked examples; nothing on this page claims to be a real client.
Where this fits
Mid-market enterprises
Organisations large enough to have real system complexity, small enough that a global consultancy's engagement model does not fit the budget or the pace.
Post-pilot organisations
Teams that ran an AI pilot, saw it work in a demo, and cannot get it into production because the platform underneath was never built.
Regulated sectors
Finance, healthcare, logistics and manufacturing, where governance and auditability are engineering requirements rather than paperwork.
Have a problem that looks like one of these?
And if it does not, that is the usual case rather than a problem — the map above is wider than the three written up here.
The question, from accounts payable
“Invoice INV-88214, €48,900 from Rehder GmbH. It doesn’t match the PO. Can I post it?”
- No source. Which policy says 3%? Nothing to open.
- One system. It never looked at the other two ERP instances.
- Not reproducible. Ask next quarter, get a different tolerance.
- Nothing for an auditor. No record this was ever asked.
- Confidently wrong reads exactly like confidently right.
- resolving identity · accounts-payable, DE entity
- retrieving · three-way match · tolerance · goods receipt
- querying · 3 ERP instances for INV-88214
- 4 passages retained of 21 · threshold 0.71
- AP Manual v6.1 · §4.3 · revised 2026-05-02Price variance is auto-cleared to 1% or €250, whichever is lower. Anything above requires the originating buyer’s approval.
- ERP query · 3 instances · liveINV-88214 is present in NAV-DE and SAP-AT. The Austrian instance already shows it as parked.
The question, from an HR business partner
“Employee in Germany is asking about parental leave. What are they entitled to, and do we top it up?”
- Statute and policy blurred. The law is right; the company part is invented.
- No source. No works agreement, no clause, no date.
- No jurisdiction check. It never confirmed which entity employs them.
- Nothing for an auditor, and this one ends up in an employment file.
- Confidently wrong reads exactly like confidently right.
- resolving identity · HR business partner · DE entity
- retrieving · Elternzeit · Elterngeld · company supplement
- separating · statutory sources from internal policy
- 3 passages retained of 17 · threshold 0.71
- BEEG §15 · statutory · consolidated 2026-01Entitlement to up to three years of Elternzeit per child, of which 24 months may be taken between the third and eighth birthday.
- Betriebsvereinbarung 2024-07 · §9Covers unpaid leave scheduling and return-to-work rights. No pay supplement is defined.
The question, from a data lead
“Can I send our customer list to the new analytics vendor for the pilot?”
- Legally wrong, and fluent about it. Pseudonymised data is still personal data.
- Invented threshold. There is no pilot exemption from Art. 28.
- No vendor check. It never looked at whether a DPA exists.
- Nothing for an auditor — the one answer here a regulator would ask to see.
- Confidently wrong reads exactly like confidently right.
- resolving identity · data lead · EU entity
- retrieving · processor agreement · transfer · pseudonymisation
- checking · vendor register for a signed DPA
- 3 passages retained of 12 · threshold 0.71
- GDPR Art. 4(5) and Recital 26Pseudonymised data that can be attributed to a person using additional information remains personal data.
- Vendor register · queried liveNo processor agreement is recorded for this vendor. Status: not approved.
The question, from a support manager
“Customer says we breached the four-hour response SLA twice this month. Do we owe them credits?”
- No contract read. It never checked which tier this customer is on.
- Invented figures. The 5% and the threshold came from nowhere.
- No clock definition. Response time from ticket creation, or from triage?
- Nothing for an auditor, and credits are a revenue adjustment.
- Confidently wrong reads exactly like confidently right.
- resolving identity · support manager · EMEA
- retrieving · service credits · response SLA · tier
- checking · this account’s contract tier
- 4 passages retained of 19 · threshold 0.71
- MSA Schedule 2 · Business tier · signed 2025-11Response target four business hours. Credits accrue from the third breach in a calendar month, at 2% of monthly fee per breach.
- Ticket log · queried liveTwo breaches recorded this month. Both measured from first human response.
- Sources shown. Both passages above, with section and revision date.
- Pinned. Model version, prompt version and index build recorded.
- Reproducible. Same inputs, same retrieval, same answer — replayable.
- Audit record written. Who asked, what was read, what was said.
- Says what it does not know, instead of filling the gap.