Skip to main content
Work

Real business challenges.Built into working systems.

The work starts inside the business. We look at the process, the people, the data, the systems and the decisions around the problem, then build what needs to exist for the operation to work better.

Scroll through the work
Expertise at scale
01 / 03

A complex recommendation depended on knowledge held across several specialists.

Every recommendation sat at the intersection of product knowledge, customer context and specialist domain expertise. As the product range, customer base and professional knowledge grew, matching each customer with the right solution became harder to scale.

18Specialists holding critical expertise
Days → secondsFrom waiting for specialist input to an immediate answer
HundredsSpecialist hours released each month
3–5Experts approaching retirement
The business challenge

Three kinds of expertise had to meet before a recommendation could be trusted.

The sales team needed deep knowledge of a growing product portfolio and the differences between available options.

They also needed to understand the specific customer: previous purchases, operating environment, past issues, preferences, constraints and what mattered in the current decision.

A third layer came from specialist domain expertise. Someone had to understand the customer’s real operating need well enough to translate it into the technical characteristics the product needed to deliver.

The strongest specialists had accumulated years of knowledge that was difficult to transfer. Their availability was limited, answers could take days or weeks, several were approaching retirement, and their knowledge did not always keep pace with new research and developments.

What we designed

A decision system where each expertise layer has a clear role.

A Product Agent understands the catalogue, specifications and differences between available solutions.

A Customer Agent brings the account context: previous purchases, history, preferences and past issues.

A Domain Expert Agent understands the operating environment and translates the need into the technical characteristics the solution should meet.

An Orchestrator receives the question, determines which expertise is required, activates the relevant agents and combines their input into one coherent response.

What this creates

Expertise becomes available at the moment the decision is being made.

The salesperson asks the question in the same language they would normally use with an experienced colleague. The system retrieves the relevant context, consults the appropriate expertise and produces one recommendation with the reasoning behind it.

Human specialists move into a validation role. Critical organizational knowledge becomes easier to preserve, update and apply across the wider sales organization.

A better recommendation, built from more of the company’s knowledge.

An agent can work across decades of product history, customer context, technical knowledge and accumulated company experience at once. A salesperson who has been in the role for two years cannot realistically carry the full depth of a 50-year organization in memory.

The result is a recommendation that is more complete, more precise and better matched to the customer’s actual need.

  • Quality of answer
  • Match to an experienced salesperson’s recommendation
  • Match to the domain expert’s recommendation
  • Recommendations constrained to valid, existing products
Decision architectureThree expertise layers → one recommendation
Expertise 01Product knowledge

Catalogue, specifications, technical differences and valid product options.

Expertise 02Customer context

History, prior purchases, issues, preferences and current constraints.

Expertise 03Domain expertise

Operating need, professional knowledge, research and technical requirements.

OrchestratorDetermine what expertise is needed. Bring it together.
Working outputRecommended solution + reasoning

Built from the customer context, valid products and the specialist logic required for the decision.

Operating efficiency
02 / 03

The workforce was expensive. The business could not see what it was really getting in return.

The immediate concern was cost. The deeper challenge was that the organization had never connected the operational, workforce and commercial signals required to understand real contribution.

30%Lower workforce cost
43%Higher productivity per payroll cost
5–6Connected data sources
3–4 daysManual analysis removed each month
The business challenge

The data existed. The relationships between the data had never been used to manage the problem.

The business had sales data, working hours, shifts, targets, employee-reported activity, store performance and a management hierarchy.

Each source was being looked at largely on its own. That left the organization unable to answer the questions that actually determine workforce value.

Did sales happen while the employee was working? Was performance driven by the employee or by a strong location? Did reported activity align with actual commercial results? Were recurring problems being acted on by management?

The manual process also required several people to check and reconcile information, creating room for human error.

What we found

The insight was hidden in the connections.

Once working hours are connected to sales, productivity becomes measurable. Once reported activity is compared with actual results, anomalies become visible.

Once individual performance is compared inside the same store and over time, the business can separate a strong location from a strong employee and identify recurring patterns.

Once the management hierarchy is connected to those patterns, the organization can see whether performance issues are actually being addressed.

The individual datasets were not the insight. The insight came from connecting them.
What we designed

One performance model across workforce, sales and operations.

The solution brings together working hours and shifts, sales data, targets, reported activity, store context and organizational structure.

On top of that foundation, the analytical layer evaluates productivity, unusual behavior, reporting reliability and performance patterns continuously.

  • What happened while the employee was actually there?
  • How productive were the hours the business paid for?
  • Is performance driven by the person or the location?
  • Is reported activity consistent with underlying data?
  • Is the anomaly isolated or recurring?
  • Did management act when the pattern became visible?
What this enables

Better staffing decisions start with a better understanding of contribution.

Management can identify productivity differences, recurring anomalies, ineffective staffing, unreliable reporting, unrealistic targets and management gaps.

Historical sales patterns, shifts and location performance can then support smarter staffing recommendations: how many people to deploy, where to deploy them and at what times.

The operating goal is clear: understand where people create value, why they create it, and where paid capacity is not producing the expected result.

Performance intelligenceUnified workforce view
SoulMynd.aiEnterprise Workspace
Last 30 daysAll locationsT
Operating performance

Sales Workforce Performance

Connected view across workforce, sales and location data.

Live performance view
Workforce cost−30%Lower payroll burden
Productivity / payroll cost43%Higher productivity
Performance review30%+Low performers identified
Monthly analysis3–4 daysManual work removed
Productivity by advisorPerformance trend
Advisor 014
88
Advisor 027
34
Advisor 031
94
Advisor 042
58
Sales / paid hour+18%
Advisor performanceView all
AdvisorProductivityReliabilitySales /
hour
Status
A14Advisor 014
88VerifiedAbove baselineHigh performer
A27Advisor 027
34ReviewBelow baselineNeeds review
A31Advisor 031
94VerifiedAbove baselineHigh performer
A42Advisor 042
58MismatchInconsistentReliability issue
Customer intelligence
03 / 03

The business had thousands of conversations. Two people were trying to listen to them.

Manual review was part of complaints, escalations, quality routines and onboarding. It consumed almost two full-time roles and still exposed the organization to only a small sample of what customers were actually saying.

3,000Calls every day
2Near-full-time reviewers listening manually
100%Meaningful-call coverage
95%Routine manual listening reduction
The business challenge

A valuable source of customer knowledge was trapped inside recordings.

Two dedicated team members listened when a customer complained, when a case escalated, during ongoing quality routines and when a new representative needed to be evaluated.

Every review required someone to find the conversation, listen to it, understand what happened and decide whether the representative handled it correctly.

Despite that effort, the organization could still examine only a fraction of the conversations taking place.

Signals of churn, recurring service problems, competitor mentions, repeat-contact reasons, sales behavior and script compliance were all present in the conversations, but difficult to use at scale.

What we saw

The calls were a business data source.

Once conversations can be understood systematically, the organization can define the signals that matter and measure them continuously across the conversation base.

Which calls should somebody listen to? becomes: What are all of our customer conversations telling us?
What we built

Vocalis. Conversation intelligence plus query agent.

Vocalis turns conversations into structured information and lets the organization define the KPIs it wants to monitor: churn signals, script compliance, competitor mentions, complaints, repeat contact and any other signal that matters to the operation.

Before analysis, the system can filter calls where there was no answer, an immediate disconnect or no meaningful conversation, keeping the analytical layer focused on interactions that contain useful information.

Managers can then move from a KPI into the conversations behind it and continue with direct questions against the call data.

From listening to learning

Human judgment moves to the conversations that deserve it.

The system surfaces patterns, anomalies and interactions that require attention. Managers gain a broader view across customers and teams, while reviewers focus on the cases where human judgment creates the most value.

Every meaningful conversation becomes part of a growing source of organizational knowledge.

Conversation intelligence modelFrom manual listening to full meaningful-call analysis
Conversation volume 3,000
calls / day

Large-scale customer interaction creates more signal than a manual review team can realistically absorb.

Complaints
Escalations
Quality review
New reps
Signal preparation

Only meaningful conversations enter analysis.

No answerExcluded from analysis
Immediate disconnectExcluded from analysis
No real conversationExcluded from analysis
Real customer interactionSent into Vocalis
Before

2 manual reviewers

Listening work consumed almost full-time capacity and still covered only a sample of the conversations.

CoveragePartial sample
MethodListen one call at a time
VisibilityFragmented
With Vocalis

Meaningful conversations become analyzable data.

The organization gains continuous visibility across the conversation base, without relying on routine manual listening.

Coverage100% meaningful calls
Routine manual listening−95%
Manager accessKPI dashboard + query agent
What the business can work with
KPI dashboardTrack the signals that matter to the client.
Churn detectionIdentify conversations that indicate likely customer loss.
Script complianceSee whether representatives followed the expected flow.
Query agentAsk direct questions against the conversation base.
Vocalis product interface
Q01Did this representative follow the required conversation script?
Q02Which stages of the script are representatives most frequently skipping?
Q03How many times were competitors mentioned this month?
Q04What did customers complain about last week?
Q05Break conversations about billing and payments into subtopics and show the statistical distribution.
Q06What are the three most common reasons customers contact us again within three days?
Across the work

The same discipline applies across very different business problems.

01
Understand the operation deeply.

Start with what is actually happening inside the business.

02
Find the constraint that matters.

Separate the visible symptom from the underlying business problem.

03
Connect the information required for the decision.

Bring together the context, data and knowledge that were previously separate.

04
Design around the people and systems already inside the organization.

The system has to work where the real operation happens.

05
Build the technology required to make the new way of working real.

Agents, data products, workflows, automation, internal applications, integrations or custom software.

Start with the challenge

Bring us the business challenge.
We'll build what makes it work.

Tell us what is getting in the way, what needs to improve or what the business needs to achieve. We’ll get inside the operation, understand what is really happening and build the system that moves it forward.

Bring us a challenge