Put your ChatGPT subscription to work with Hyperagent

OpenAI just launched Sign in with ChatGPT, which allows Plus and Pro subscribers to use tokens from your ChatGPT subscription in products like Hyperagent. Here's what customers are achieving with OpenAI's models in Hyperagent.

Appell Striping: $2M in sales added with agents

Michael Appell runs a national parking-lot striping and sealcoating business with a network of franchisees. His agents find prospects, prepare targeted outreach and pass qualified replies to the local franchisee.

Across the national division, Michael reported 48% more approved jobs, a 56% increase in their value and 70% more gross profit on approved work, compared with the previous year. He credited agents with helping his team send quotes faster and obtain more quotes for each job.

"I've never had that type of growth before, and we've been doing this for 25 years, 26 years now."Michael Appell

He had started without coding experience. His projects now include the parking-lot counter, a condition-report tool that franchisees review and correct on site, training videos and territory-mapping software. The ambition grew from one useful job into a set of tools for the business.

Sales prospecting: Showing the before-and-after transformation

Michael's agents are given a set of ZIP codes to prospect. Agents use satellite and streetview imagery to identify commercial properties with large parking lots that are due for maintenance.

For each prospect, the outreach agents generate "before and after" images that pair the aerial image of the actual lot with a generated image showing fresh sealing and striping. Using this strategy, Michael's response rates to cold outbound have jumped over 3%.

Michael reviewed the emails for the first few weeks before allowing the agent to send without his approval on each message. The agents now handle the repeated research, preparation and routing. Franchisees bring their knowledge of the local market and build the customer relationship.

Quoting: Complex visual reasoning

When a prospect responds with interest, the franchisee visits the property, explains the work and prepares a quote.

To prepare a striping estimate, a franchisee needs to know how many spaces the job covers. Michael's counter reads an aerial image and returns the number of parking stalls and accessible spaces. In his demo, it reports 89 stalls and six accessible spaces. That replaces the first manual count with a starting point for the quote.

Building around the strengths of each GPT model

Here's how customers like Appell are designing their agents around the unique strengths of the GPT model family.

1. Researching prospects in parallel: GPT-6 Luna. Turn search results into company profiles, contact lists and market maps. Luna's fast, lightweight processing suits high-volume research and regular updates across many records.

2. Complex analysis and planning: GPT-6 Astra. Compare suppliers, evaluate market opportunities or combine evidence from several sources into a recommendation. Astra suits work that needs deeper reasoning across multiple steps.

3. Visual concepts and transformations: GPT Image 2.5 Sunburst. Turn product or property photos into visual proposals, campaign concepts and customer-facing assets. Sunburst combines reference-image editing with high-fidelity output to show what a proposed result could look like.

4. Personalized communication and reply routing: GPT-6 Luna. Turn customer records and briefs into email drafts, summaries and responses. Classify incoming messages and route them to the right person or next step.

5. Business tools and workflow automation: GPT-6.1 Sol. Build quote calculators, reporting dashboards and review tools. Sol's coding and tool-driven reasoning support the work of building, testing and refining software for the team.

More teams putting agents to work

Other teams use agents to find customers, research startups and turn customer feedback into campaign ideas.

HelloPackage

Reach a new market faster

HelloPackage used a prospecting agent to cover 27 markets and identify approximately 800 buildings in one day. CEO Paul Favorov said the work had previously taken months of manual research. The result was a new set of potential customers for his team to pursue.

"What really made it hit for me is how quickly, within a day, it did about 27 markets and identified about 800 opportunities for us to go after."

Paul Favorov, CEO

HearstLab

Research thousands of startups

At HearstLab, an agent named Ivy finds and researches startups, prepares recommendations and drafts messages to founders. Dominique, its Data & Operations Director, reports 4,500 startups processed and 2,300 hours saved for the team. People still make the investment decisions.

"It's not something a human could do."

Dominique, Data & Operations Director

Winona

Turn customer reviews into campaign ideas

Winona uses agents to collect Trustpilot reviews each week and develop customer profiles and campaign ideas from what people actually say. A strategy agent passes ideas to a copywriting agent, and a person reviews the copy. Chief Marketing Officer Rob Schmidt describes eleven or more named agents working in the business.

"We can start to speak to individuals right where they are, in a much more scalable, essentially programmatic way."

Rob Schmidt, Chief Marketing Officer

For an agency or a small marketing team, the appeal is easy to recognize: less time assembling the research, more time deciding what to say.