This blog post was inspired by a conversation we had with a client that is currently going through a company-wide AI transformation and exploring ways to accelerate.

AI transformation doesn’t happen on its own. Giving ChatGPT (or any other tool) to a handful of enthusiasts creates the illusion of progress, but it rarely changes how the firm actually works. If anything, it breeds complacency: we’ve got AI covered. Meanwhile, competitors that go all‑in are compounding advantages in speed, coverage, and cost.

This post outlines a pragmatic, management‑led approach we’ve seen work across Quantly’s customers: universal access, hard ownership, measurable adoption, and a steady march from general‑purpose tools to domain‑specific agents and internal builds.

The False Comfort of Pilots and Power Users

Small pilots are appropriate for de‑risking, but they are not a transformation strategy. Typical failure modes:

• Visibility without velocity. A few demos and sporadic wins, but no change in desk‑level throughput or efficiency.

• Tool sprawl. Every team experiments with something different, creating governance and security headaches. The lack of an intentional choice of partner also hinders progress into internal data integrations and embedding AI deeper into the business.

• Pockets of automation. Early AI adopters create local pockets of value while the average analyst’s workflow remains unchanged.

• Limited experimentations. Limiting AI access to a few power users restricts experimentation and innovation, as their personal biases dictate tool usage. Broad access, combined with a mandate to use the tools, unlocks a wider range of ideas, use cases, and workflow improvements, driving benefits across the entire organization.

If the goal is competitive advantage, the bar is higher: the average analyst must be able to cover more companies, spot more opportunities, and generate differentiated insights faster, every day.

Make It a Mandate: Top‑Down and Firm‑Wide

AI transformation is an operating change, not a software rollout. Treat it like any other strategic program:

1. Executive mandate. Set a public goal (e.g., “By quarter‑end, every desk will have X workflows automated by AI”). Make leadership visibly use AI in their own routines.

2. Universal access on Day 1. Enable a general tool (e.g., ChatGPT or your approved equivalent) for everyone to remove the access bottleneck.

3. Move quickly to specialization. Within a quarter, introduce domain‑specific tools and internal builds that plug into analyst workflows (deep dive research, company meeting prep, news monitoring, earnings previews etc). General chat is the on‑ramp; specialized agents are the highway.

4. Measure, publish, improve. Track usage and outcomes by team. Iterate on feedback. Share league tables. Celebrate wins and remove blockers.

Segment Your Users and Plan the Interventions

Rolled out to hundreds of users, you’ll consistently see three cohorts. Design for each:

• Early Adopters: Daily users from the start. They understand LLMs’ strengths/limits and push boundaries.
  What they need: access to advanced features, a voice in shaping the AI roadmap, and a channel to showcase their agents and playbooks to others.

• Curious Explorers: Dip in and out to see what’s possible. Open‑minded but not yet habitual users.
  What they need: nudge loops, quick‑wins training, and curated prompts/agents directly tied to their workflows so they can easily take and adopt.

• Skeptics: Have tried AI, were disappointed, and moved on.
  What they need: structured, manager‑led trials on specific tasks, workshops that show rather than tell, side‑by‑side comparisons to baseline and visible peer evidence.

Rule: Don’t wait for culture to evolve organically. Engineer it with targeted enablement and accountability by cohort.

Assign an Owner To Bring AI Into Teams

Name an AI Transformation Lead (program owner) who sits between the business and tech. This person should have domain expertise, ideally someone with an analyst background that is able to understand and relate to analyst peers and convey the value of the AI initiative. Their remit:

• Intake & discovery. Go team by team; map recurring workflows that lend themselves to augmentation or automation.

• Prioritization. Build a queue of use cases with clear ROI, compliance posture, and data needs.

• Delivery. Configure/build agents or playbooks; map sources of data needed to integrate; deploy into teams; document; organise workshops and training.

• Adoption. Set weekly usage goals with team leads; publish dashboards; run office hours.

• Iteration. Collect feedback, improve playbooks and agents based on analyst feedback, retire what doesn’t work.

If necessary, back them with an exec sponsor, a small enablement squad, and access to engineering for connectors/integrations.

From General Tools to Purpose‑Built Agents

A sequencing that works:

• Stage 1: General AI for everyone. Baseline familiarity, firm‑wide prompts library, light guardrails. This mostly involves onboarding a general AI tool such as ChatGPT, Claude or Microsoft Co-pilot. At this point users will start getting familiar with the technology and start adopting it for human-centric use cases (e.g. writing up an email), but will realise that general tools may not be sufficient to help them with their domain-specific analyst work.

• Stage 2: Introduce domain-specific tools that meet the bar for accuracy, domain knowledge and workflow fit. These tools should have a high degree of customisation for specific analyst workflows and move beyond chat, to more involved end-to-end workflows that can connect with your internal data, systems and fit into your security and compliance requirements.

• Stage 3: Internal builds and deep integrations. Connect internal data: drives, data warehouses, and other external data providers. Build out a reliable AI infrastructure composed of a data layer and an agentic layer so that you can scale and deploy agentic use cases quickly. There are multiple intricacies with all of those and often firms may bring in a partner to build this alongside to accelerate delivery.

The goal is to move from chatting to operating and building an AI edge: reliable, repeatable workflows with auditable outputs.

Workshops, Training, and Upskilling That Actually Change Behavior

• Role‑based curricula. Different roles and teams of analysts and PMs can receive tailored training based on the work they do and the AI workflows they leverage.

• Show‑don’t‑tell sessions. Live builds of a real desk use case in 30 minutes.

• Prompts → Playbooks. Convert popular prompts into one‑click agents with inputs, actions, and outputs. Go from ad-hoc Q&A to end-to-end workflows.

• Office hours + helpdesk. Daily 30‑minute drop‑ins; Communication channels for feedback and idea sharing. Implement feedback to show involvement and engage analysts further.

• Manager check‑ins. Adoption is a standing agenda item in weekly desk meetings.

Common Objections and How to Respond

• “Quality is inconsistent.” Move from free‑form chat to hyper-customised, yet opinionated agents with source attribution, tests, and guardrails. Aim to design agents with formatted outputs. Establish a review loop like you would for any analyst output.

• “Security/compliance risk.” Use enterprise authentication, role policies, and full audit logs. Start on lower‑risk workflows; expand as controls mature. Depending on internal security policies, opt for a partner that is comfortable deploying in your environment to fit into your firm’s existing security perimeter.

• “We’ll wait for the tech to settle.” The tech is far from settling and waiting is falling behind. Competitors are learning and building capabilities now; each month of delay widens the gap and makes adoption harder later.

• “We tried ChatGPT; it didn’t stick.” Tools aren’t the strategy. Mandate usage on named workflows with training, adoption targets, and visible leadership participation.

The Takeaway

AI transformation works when it’s deliberate, inclusive, and led from the top. Pilots and isolated wins are useful, but durable value comes from putting AI in everyone’s hands, moving from general chat to domain-specific tools, wiring agents into real workflows, and measuring adoption over time.

If you’d like to explore what this looks like for your teams, get in touch. We can share examples from our own experience of successful and ongoing partner relationships and how firms have leveraged Quantly to set the foundations for scalable and transformative AI adoption.