Building trustworthy systems in the age of AI while remaining human at heart

When I joined HubSpot, I entered a rare situation. I had already spent years as a customer learning how to creatively build systems using the tools at my disposal. I then joined the company with the task of modernizing a long-standing customer reference system that had served many teams well but was now struggling to meet new expectations, complexity and scale.

Seeing both sides changed my approach to this work. Advocacy is often misunderstood. It can be viewed as simple or administrative because much of its complexity occurs behind the scenes. But if you look closely, you realize that it requires nuance, judgment, finesse and emotional intelligence at every step.

My goal wasn’t to replace any of it. It was about creating a system that supports this.

If you’ve ever tried to build trust on a large scale, you probably know firsthand how challenging the work can be. So take a look at what we’ve rebuilt at HubSpot, how we did it, and how you can apply the same principles without needing an engineer or a separate platform. And speaking as someone who isn’t an engineer at all, just a marketer with a MacBook and guts: If I can build this, you can too.

If there was one theme on this trip, it was this AI is not a threat to fear. is inconsistency. AI has not removed the human parts of this work. It has been made clear where they are most important.

The quiet work behind every victory

Every organization relies on work that is often invisible but extremely impactful:

• The coordinator who detects a potential discrepancy before it becomes a problem.
• The specialist who remembers a customer’s context, which no system fully captures.
• The representative that adds an additional sentence that changes the quality of a match.

Advocacy teams live here every day. They create credibility, connections, and evidence in ways that are easy to underestimate when the process is scattered or opaque. As a former customer and now a HubSpotter, I’ve seen how often the work was undervalued, not intentionally, but because its complexity was hidden.

The goal of this transformation was to make this work visible, respected and supported so that people have the structure they need to excel.

AI has not replaced humans. It supported her.

As we redesigned the reference process, one thing became very clear: the system had become increasingly complicated over time. It wasn’t because the work was flawed. The people trying to help filled in the gaps manually.

The old process required 18 separate steps. After the renovation, a coherent sequence of five clear phases emerged.

The most surprising finding was how well AI meshed with human judgment. It did not eliminate the need for nuance or relationship context. It supported it.

• HubSpot workflows ensured predictable routing.
• Slack made communication immediate and visible.
• AI co-pilots helped validate eligibility and reduced manual triage.

This gave people more time to focus on the areas that only humans can do: storytelling, empathy, nuance and partnership.

From stories to systems and then to scaling

As the new system rolled out, it became clear that we were not only creating workflows, but also shaping how trust spreads throughout an organization.

When teams gain transparency about advocacy work, three things reliably happen:

1. Reciprocity increases.

When people realize how important their involvement is, participation grows organically. This was one of the strongest impulses.

2. Equity grows.

Advocates who had previously been overlooked emerged naturally through objective criteria.

3. Alignment is strengthened.

Sales, success and marketing began to operate on shared information rather than assumptions.

This change was less about tools and more about structure. HubSpot simply gave us the environment to create shared clarity.

Creating a single source of truth for trust

Step 1: Create a data-driven baseline.

One of the most persistent challenges for advocacy teams is demonstrating the impact of their work. ROI, impacted revenue, readiness forecasts, and coverage gaps are difficult to measure when the underlying data model is fragmented or maintained inconsistently.

Before we could optimize workflows or add automation, we needed a data foundation strong enough to support operational and reporting needs at scale.

To address this issue, we developed a trust readiness model that assesses:

• Relationship maturity, including tenure, prior collaboration, and mood patterns.
• Depth of product adoption based on usage data, feature-level adoption, and cross-portal behavior.
• Account health through renewal signals, support trends and lifecycle stage.
• Growth signals such as expansion opportunities, product interest and account history.
• Willingness to engage is measured through outreach responses, past advocacy participation, and customer feedback.

The design of this model was the conceptual part. The real work was implementing it in HubSpot in a way that was both reliable and scalable. This required building a complete data architecture that included:

• Contact, company, and deal level custom properties designed with strict naming conventions and data types to avoid future confusion.
• Validation rules that prevented incorrect or incomplete data entry.
• Conditional assessment logic that automatically updates readiness based on property changes, usage data, and lifecycle events.
• Workflow logic associated with each fulfillment stage ensures requests are processed in a consistent and controlled manner.
• Segmentation rules that recalculate advocacy readiness and match actionability in real time.
• Priority rules for conflicting values, stale data, and high-risk accounts.
• Dashboards built for diverse audiences, including ROI reporting for executives, velocity tracking for operations, and readiness insights for frontline teams.

The impact of this work was immediate. For the first time, we were able to quantify the impact of advocacy across all deals, measure actual coverage gaps, track readiness trends, and provide clear attribution to revenue. These insights were not previously possible because the system was not designed to support this level of precision.

Once the structure was in place, the CRM took over much of the ongoing calculation. We just had to be intentional about how we built the foundation.

Step 2: Build the operative bones.

Once the data layer was stable, we shifted our focus to operational design. During this phase, the backend architecture evolved into a functional and intuitive process for the teams that use it.

Our goal was to create a system where every action had a clear path, every outcome was measurable, and every stakeholder could see where a request stood without having to ask.

We started developing a multi-layered dashboard system with different views for executives, managers and operators:

• Leadership saw revenue impact, program coverage and strategic trends.
• Managers saw team engagement, request volume, and bottlenecks.
• Operators saw daily fulfillment stages, match rates and customer readiness.

We then created workflow chains that govern intake, routing, notifications, and completion:

• Intake workflows standardized the questions employees answered during submission.
• Routing workflows routed matching requests to the correct fulfillment path.
• Notification workflows ensured timely reminders and prevented delays.
• Closing workflows updated report properties and triggered follow-up steps.

We also created segmentation rules that filter advocates by readiness, permissions, region, product experience, and capacity to ensure an accurate and scalable match.

And we’ve developed brand templates to create consistency in outreach, customer communications and stakeholder updates, reinforcing professionalism and reducing cognitive load.

As the system grew, governance became increasingly important. We have implemented the following:

• Naming conventions for workflows, lists, views, and properties.
• Change management rules to avoid breaking dependencies.
• Audit cycles to identify unused resources or conflicting automation.
• Documentation for each piece of equipment and its purpose.

While this governance wasn’t particularly glamorous, it prevented deviations and helped ensure the system remained reliable even as request volumes increased and new team members came on board.

As time passed, something significant happened. With a clearer structure, shared visibility and a reliable process, advocacy began to be viewed not as coordination work, but as strategic work that helped influence sales, customer trust and the quality of the partnership. The system enhanced the work simply by revealing its complexity and value.

Step 3: Scale for speed, consistency, and transparency.

Trust quickly erodes when processes are slow, inconsistent, or unclear – especially in cross-functional work where many people rely on the same information to drive a business.

We knew that if we wanted to scale advocacy sustainably, the experience had to feel predictable, fair, and transparent for everyone involved. This meant developing a repeatable operating rhythm that closely aligned with HubSpot’s actual workflows.

To solve this problem, we created a structured fulfillment sequence that every request goes through:

Request → Route → Align → Activate → Frame → Fulfill

Each phase has a defined purpose, owner and outcome.

Nothing floats. Nothing is lost. Nothing depends on memory or individual preferences.

AI played the role of pattern recognition and validation, reducing the manual effort of scanning for product fit, regional targeting, business size considerations and previous advocacy. HubSpot helped orchestrate movement between stages through workflows and task distribution, meaning each step was visible, time-stamped and trackable. Humans intervened where nuance was needed, particularly around relationship context, customer readiness, and interpreting the nuances that no automation can fully understand.

When we built this system, something unexpected happened. Empathy towards the work itself increased noticeably. As teams recognized the complexity involved—the decisions, the careful wording, the balance between customer care and revenue impact—they developed a deeper appreciation for the people behind the scenes who made the process work. The system made the intricacies visible, and with visibility came greater kindness, patience, and collaboration.

To strengthen this structure, we introduced a two-person Reference Fulfillment Ops Pod:

• The Coordinator manages onboarding, triage, training, and coordination across the Slack Helpdesk.
• The specialist takes care of the more in-depth evaluation, customer approach and the connective tissue of the dating agency.
• Your work is supported by SOPs, structured views and multiple GPT co-pilots that reduce manual effort on tasks such as letter writing and reporting.

Taken together, this created a system in which most of the operational load is automated or assisted, but the remaining human decisions provide trust. In the final step, empathy, judgment and relationship management come into play. And now that the intricacies are being made visible, this work is being respected and valued in a way that it often wasn’t before.

Step 4: Redefine reciprocity and internal culture.

Systems can enable advocacy, but culture is what sustains it over the long term. A process won’t thrive if people don’t see themselves in it or if the work feels transactional. We needed a cultural foundation based on mutual recognition, shared responsibility and genuine appreciation for the emotional intelligence required to do good work.

Advocacy is not just operational. It’s relational. It requires empathy for both customers and internal teams, as well as a sense of timing, context and capacity. The more we bring these nuances to light, the better teams will understand why thoughtful engagement is important.

To reinforce this change, we drew on the principles of learning systems and group psychology. Instead of forcing participation, we modeled the behavior we wanted to encourage. We made work more transparent, shared context more proactively, and highlighted small successes alongside big ones. We have shown how advocacy is linked to customer trust, business speed and long-term relationships.

One of the most effective rituals turned out to be incredibly simple. Each quarter we recognize the representatives who have worked most actively with the program. We publicly celebrate their collaboration, recognize their managers, and recognize the impact of their efforts. The recognition wasn’t about the scoreboard culture. It was about honoring the emotional labor, judgment and relationship building that often goes unseen.

The result was a cultural change. Advocacy no longer felt like a request-based proposal, but rather a shared partnership. With greater visibility came greater empathy. Teams began to understand the intricacies involved and responded with more care, context and collaboration. Representatives participated earlier and more thoughtfully. The managers were proud of their teams’ commitment. Managers incorporated insights from advocacy into planning discussions.

Reciprocity became the cultural norm because the work finally felt understood.

The deeper truth: systems for people

Many systems track activity, but very few are designed to support people’s work. Redesigning the reference process gave us the opportunity to develop something more thoughtful. A structure that:

• Respect time.
• Award for expertise.
• Reduces friction.
• Surface contributions.
  Makes trust measurable.
• Supports work that has long been underestimated.

HubSpot provided the tools, the architecture provided clarity, and the people provided the heart and mind.

A note for the builders

If this rebuilding has taught me anything, it’s that trust doesn’t happen by accident. It is created through systems that respect the people who do their work and enable them to work with clarity, consistency and care.

What we have developed at HubSpot is just one example of what this can look like. The details vary for each team, but the underlying principles remain the same:

• Create a data basis that you can rely on.
• Create workflows that support human judgment rather than override it.
• Build reporting models that make impact visible.
• Protect the people doing the work with structure, not more effort.
• Strengthen the culture by showing what good looks like, not by enforcing it.

This case study is specifically designed for teams building within constraints. For the operators who live in CRMs and spreadsheets, trying to create order in inherited chaos. For the program managers who may not have a dedicated engineering partner or budget for a dozen specialized tools, but have access to HubSpot and a clear vision of what they want the customer experience to feel like.

You don’t need a complex tech stack to build something useful. You need clarity, thoughtful architecture, and a willingness to find solutions for people on both sides of the process. The rest can be created, improved and iterated layer by layer.

If you recognize yourself in this work, know that you are not alone. The impact you make may not always be visible, but it is measurable, repeatable and essential. And with the right system behind you, it also becomes scalable.

That is the real insight behind this conversion.