How Guidesly constructed AI-generated journey studies for out of doors guides on AWS

That is visitor publish by David Lord, Taylor Lord, Shiva Prasad, Anup Banasavalli Hiriyanagowda, Nikhil Chandra from Guidesly.

Guidesly is reshaping how out of doors recreation is booked, run, and skilled. Based in 2019, it started as a method to join anglers, hunters, divers, and out of doors recreation fanatics with trusted guides, dive retailers, and charters. It has since grown right into a vertical AI software program as a service (SaaS) system serving the whole trade. With Guidesly Professional, out of doors professionals acquire a enterprise resolution that powers each a part of their operation—bookings, funds, web sites, shopper administration, and advertising—all from a single system.

For a lot of guides, the hardest problem is getting found and slicing by the noise on-line. Even those that know what have to be finished can spend as much as eight hours a day updating web sites, posting on social media, and operating electronic mail campaigns. With out constant execution throughout these channels, visibility drops, and smaller operators danger falling behind rivals with full advertising team-missed alternatives that immediately affect development and bookings.

It was addressing this drawback that Jack AI was born. From the beginning, Guidesly noticed AI not solely as a software, however as a method to join the silos that guides face day by day—uniting bookings, knowledge, content material, shopper engagement, and advertising into one clever circulation. The imaginative and prescient went past automation. It was about creating a real accomplice that works alongside guides, quietly dealing with the heavy lifting they don’t have time for.

In contrast to general-purpose AI instruments that require fixed prompting and oversight, Jack AI works within the background by itself. It prompts robotically after every journey, remodeling uncooked knowledge, pictures, and movies into polished, ready-to-publish content material throughout web sites, social media, and electronic mail. Operating serverless on AWS, it scales robotically to ship constant content material at pace, permitting guides to give attention to their journeys somewhat than administrative work.

On this publish, we stroll by how Jack AI is constructed on AWS to energy this end-to-end automation. We discover how providers resembling AWS Lambda, AWS Step Features, Amazon Easy Storage Service (Amazon S3), Amazon Relational Database Service (Amazon RDS), Amazon SageMaker AI, and Amazon Bedrock come collectively to ingest journey media, enrich it with context, apply laptop imaginative and prescient and generative AI, and publish marketing-ready content material throughout a number of channels—securely, reliably, and at scale.

The problem: Releasing guides from advertising operations

For out of doors guides, the actual aim is delivering really memorable experiences, however creating partaking content material stays a vital and time-consuming process. Every journey produces dozens of pictures and tales, but turning them into compelling advertising is a problem:

• Figuring out species and journey particulars – Guides seize numerous pictures, however manually tagging species, sizes, strategies, and places is painstaking. Lacking particulars could make posts much less informative and fewer partaking for potential purchasers.
• Capturing the proper voice – Each information has a singular type formed by native jargon, private storytelling, and years on the water or within the discipline. Writing content material that feels genuine—with out sounding generic or mismatched—is sort of unimaginable to scale.
• Maintaining with Search engine marketing – Persistently producing keyword-rich, domestically improved content material is difficult even for skilled entrepreneurs. For busy guides, missed Search engine marketing alternatives imply decrease discoverability and fewer bookings.
• Managing a number of channels – Journey report pages, blogs, Instagram captions, Fb posts, and electronic mail newsletters all demanded consideration. Updating these assets manually meant hours of writing, modifying, and formatting each week.
• Sacrificing time on the water – Each hour spent at a laptop computer is an hour taken away from guiding purchasers. For small companies, this tradeoff impacts each income and buyer expertise.

Even with guides doing their greatest, handbook processes weren’t correct or quick sufficient to maintain tempo with shopper demand, trendy advertising wants, and the vital want to remain related with clients through electronic mail. That’s the place Guidesly’s Jack AI steps in—automating content material creation, Search engine marketing optimization, electronic mail advertising, and multi-channel distribution, so guides can give attention to what they love: delivering unforgettable out of doors experiences.

Overview of the answer: Jack AI

To carry Jack AI to life, Guidesly applied a completely automated, serverless, AI-driven advertising workflow on AWS, designed to remodel uncooked journey knowledge into ready-to-publish content material. This technique permits guides to give attention to delivering distinctive out of doors experiences whereas sustaining a constant, genuine digital presence throughout web sites, social media, and electronic mail campaigns.

The next diagram illustrates our pipeline:

Journey media ingestion – Amazon API Gateway:

1. Computerized set off: Journey pictures and movies uploaded by guides to enter the system by Amazon API Gateway, which instantly triggers the orchestration pipeline.
2. Contemporary content material supply: Media is processed as quickly because it’s uploaded, enabling social posts and electronic mail campaigns to achieve audiences whereas journeys are nonetheless high of thoughts.

Pipeline orchestration

AWS Step Features handle the workflow, invoking AWS Lambda features for every stage—from knowledge extraction and fish detection to media enchancment, content material technology, and publishing.

1. Information extraction from media
   1. Computerized metadata seize: As quickly as guides add pictures and movies, the system extracts embedded EXIF Metadata, together with GPS coordinates, timestamps, and machine settings.
   2. Contextual enrichment: The extracted geospatial data is then mixed with related climate and water situation knowledge for a similar time and placement. This captures particulars resembling tide ranges, water temperature, wind pace, and cloud cowl—context that may in any other case be misplaced.
   3. Richer storytelling: By grounding every journey within the precise environmental situations alongside what was caught, the system produces content material that’s extra personalised, genuine, and interesting, with out requiring extra effort from the information.
   4. Scalable consistency: Whether or not processing a single picture or a whole lot, automation makes positive that each media artifact is enriched with high-quality contextual knowledge, offering dependable inputs for downstream processes.
2. Fish species detection utilizing laptop imaginative and prescient

Fish species identification is among the core capabilities of the Jack AI system. The problem isn’t solely detecting fish in real-world photographs but additionally precisely classifying a whole lot of species throughout extremely variable environments resembling boats, docks, lakes, and offshore places.

To deal with this, we designed a multilayer laptop imaginative and prescient pipeline that mixes custom-trained laptop imaginative and prescient fashions with basis imaginative and prescient fashions out there by AWS providers.

Experimentation and mannequin improvement

Our mannequin improvement workflow runs primarily inside Amazon SageMaker AI utilizing JupyterLab because the experimentation setting.

With this setup, we are able to:

• Quickly prototype new laptop imaginative and prescient architectures
• Run large-scale coaching jobs on GPU-backed cases
• Consider fashions throughout a number of fish classification benchmarks
• Iterate shortly between mannequin enhancements and manufacturing deployment

The SageMaker AI setting acts because the central hub the place datasets, coaching scripts, and mannequin experiments are managed.

Dataset and coaching challenges

Fish identification presents a singular machine studying (ML) problem because of the massive variety of species and uneven knowledge distribution. Our system at the moment helps over 400 fish species courses, collected from a mixture of:

• Proprietary fishing report imagery from the Guidesly system
• Consumer-submitted catch pictures
• Curated datasets gathered from accomplice sources

Whereas some widespread species have hundreds of coaching examples, many species have restricted labeled photographs, which makes conventional supervised studying approaches tough.

To deal with this imbalance, we use a hybrid coaching technique:

• Customary supervised studying for species with massive datasets
• One-shot and few-shot studying strategies for uncommon species the place coaching knowledge is proscribed

This permits the system to broaden classification protection with out requiring massive datasets for each species.

Multi-layer imaginative and prescient pipeline

Slightly than counting on a single mannequin, we applied a two-stage imaginative and prescient structure that separates object detection from species classification.

Detection Layer

The primary stage makes use of YOLO-based object detection fashions skilled to determine related objects inside fishing photographs, together with:

• Fish
• Fishing gear
• Individuals
• Boats and environmental context

The detection fashions determine bounding containers for every object. As a substitute of passing the whole picture to the following stage, we crop solely the detected fish areas.This strategy considerably improves classification accuracy as a result of it removes unrelated background parts that may confuse classification fashions.

Classification Layer

Every cropped fish picture is then handed right into a specialised classification mannequin.Over the course of improvement, we experimented with a number of architectures together with:

• Convolutional neural networks (CNNs)
• ResNet-based fashions for robust baseline classification
• One-shot and few-shot fashions for long-tail species recognition

The mixture of architectures permits us to stability accuracy, inference pace, and coaching effectivity throughout a whole lot of species courses.

Hybrid imaginative and prescient + basis mannequin strategy

Along with our custom-trained fashions, we combine multimodal basis fashions (FMs) out there by Amazon Bedrock to offer extra reasoning and contextual understanding.

Nevertheless, uncooked imaginative and prescient fashions can typically hallucinate or misread visible scenes. To cut back this danger, we apply a number of preprocessing steps earlier than sending photographs to basis fashions:

1. Picture preprocessing
   1. Cropping detected fish areas
   2. Normalizing picture dimensions
   3. Eradicating pointless background
2. Context enrichment
   1. Media metadata (location, water physique, time)
   2. Recognized species distribution
   3. Detection mannequin outputs
3. Structured system prompts
   1. Present the mannequin with contextual details about the picture
   2. Constrain potential species predictions

With this hybrid strategy, we are able to mix the precision of domain-specific classifiers with the reasoning capabilities of huge imaginative and prescient fashions.

From analysis to manufacturing

After the fashions are validated, we deploy them utilizing managed endpoints on Amazon SageMaker AI. This allows:

• Actual-time inference on uploaded photographs
• Computerized scaling for big volumes of media
• Steady monitoring and mannequin updates

The result’s a scalable imaginative and prescient system able to processing hundreds of fishing photographs throughout Guidesly’s system, delivering dependable fish species detection even in complicated real-world situations.

1. Media enchancment for quicker, web-ready publishing

After fish detection and contextual enrichment are full, Jack AI focuses on getting ready media for real-world publishing. Excessive-resolution pictures and movies uploaded by guides are robotically processed into optimized, web-ready property designed to be used throughout web sites, social networks, and electronic mail campaigns. This enchancment pipeline handles compression, resizing, and format conversion behind the scenes. This makes positive that media recordsdata stay light-weight with out sacrificing visible high quality. By standardizing property early within the workflow, Jack AI removes the necessity for handbook picture modifying and maintains constant presentation throughout gadgets and programs.

Improved media is saved as versioned artifacts in an Amazon S3 bucket and tagged for easy retrieval and reuse. These property may be surfaced repeatedly throughout Search engine marketing pages, journey studies, newsletters, and social posts with out reprocessing, retaining the publishing pipeline quick and environment friendly. Past efficiency, this step additionally helps Search engine marketing objectives—fast-loading photographs to enhance search rankings, improve consumer expertise, and cut back bounce charges on information web sites.

1. Tone enchancment

To make it possible for generated journey studies really feel genuine and aligned with the pure voice of fishing guides, a Tone Enchancment layer was launched inside the content material technology pipeline. Slightly than modifying the underlying language mannequin by fine-tuning, the system improves tone by contextual inputs and structured prompting. This preserves the distinctive storytelling type guides use whereas sustaining scalability and operational simplicity.

The inspiration of this strategy is context injection. Structured journey metadata is embedded immediately into the mannequin immediate, giving the mannequin the grounded context it must generate correct and related narratives. Alongside this, historic journey studies and guide-specific phrasing patterns are retrieved and included as reference examples. This helps the mannequin mirror the vocabulary, pacing, and descriptive type that guides naturally carry to documenting their journeys. Slightly than coaching a {custom} mannequin, rigorously designed prompts information the muse mannequin towards outputs that mirror the anticipated tone. This permits for dynamic changes to writing type with out the operational overhead of sustaining fine-tuned fashions. To keep away from fabricated particulars, the technology course of is constrained strictly to the supplied metadata and contextual inputs. The mannequin is instructed to not infer lacking data like extra species, strategies, climate, or places absent from the supply knowledge, so each report stays according to the precise journey file.

Era itself is executed utilizing Amazon Bedrock FMs, which course of the contextual inputs and structured prompts to supply coherent, domain-appropriate studies at scale. By counting on contextual prompting somewhat than mannequin re-training, the system avoids coaching infrastructure, reduces operational overhead, and permits speedy iteration as new information studies and area patterns emerge. This strategy struck the stability that the system wanted: genuine, guide-style journey studies delivered with reliability, price effectivity, and the scalability to develop because the system expands.

Publishing pipeline

After the content material is generated, improved, and refined for tone, the publishing pipeline brings all the things collectively to ship marketing-ready property throughout channels. This stage is designed to run end-to-end with minimal handbook effort, ensuring that guides keep knowledgeable whereas automation handles execution behind the scenes.

Asset technology is dealt with by an Property Era Step Operate that orchestrates a number of AWS Lambda runs. These features generate advertising deliverables from the artifacts saved within the S3 bucket for every journey. This consists of Search engine marketing-friendly journey studies, recent web site content material, social media posts, and personalised electronic mail campaigns. The outputs are robotically saved within the system and built-in into downstream publishing workflows, lowering the necessity for handbook drafting, copywriting, or formatting. After the property are prepared, guides obtain push notifications for overview, to allow them to keep knowledgeable with out pointless operational overhead.

Processed artifacts—together with improved media, extracted journey particulars, and generated advertising property—are saved centrally utilizing Amazon RDS and Amazon S3. Amazon S3 supplies sturdy, cost-effective storage for media and generated content material, whereas Amazon RDS makes structured journeys and information knowledge out there for downstream workflows and reporting. Collectively, these providers make it possible for property are instantly reusable throughout web sites, social channels, and electronic mail campaigns with out requiring extra processing.

Publishing controls stay versatile by AI-driven automation. Guides can overview and approve generated content material, request refinements, or depend on a built-in auto-publish toggle for full automation. With this flexibility, every information can stability high quality management with effectivity—remaining hands-on when wanted or choosing a set-and-forget strategy. Behind the scenes, AWS Step Features orchestrate a number of AWS Lambda operate runs, scaling robotically to accommodate a whole lot of guides with minimal infrastructure administration.

Price issues

Whereas the structure is designed to scale robotically, the price per generated journey report stays comparatively small. In typical eventualities, producing a full report—together with media processing, laptop imaginative and prescient inference, and content material technology—prices roughly $0.10 to $0.50 per report. The ultimate price varies relying on elements such because the variety of photographs processed, the presence of video media, and the quantity of AI inference requests. As a result of the workflow is serverless and event-driven, guides solely incur prices when studies are generated, retaining the unit economics predictable as utilization grows.

Impression on out of doors recreation advertising

With Jack AI working end-to-end on AWS, the affect extends past automation and into how out of doors recreation advertising is executed day-to-day. By combining AI-driven automation with AWS providers, the method of producing, refining, and publishing advertising content material has been lowered to a single, repeatable workflow. Outside recreation guides now not must spend hours drafting journey studies, formatting photographs, or scheduling social posts. As a substitute, these duties are dealt with robotically, relieving guides to give attention to what issues most: their purchasers and the out of doors expertise itself.

The result is a constant, high-quality digital presence throughout web sites, social media, and electronic mail campaigns. Guides enhance visibility, strengthen search rankings, and interact clients extra successfully with out the necessity for devoted advertising employees.

Outcomes

Since launching Jack AI on AWS, Guidesly has seen speedy adoption and measurable affect throughout its neighborhood of out of doors guides. By automating probably the most time-consuming elements of their work, advertising, Jack AI has lowered the operational effort required after every journey whereas serving to guides keep seen and aggressive on-line.

Beforehand, guides typically spent greater than six hours each week behind a laptop computer writing journey studies, formatting weblog posts, creating social media captions, and making an attempt to enhance content material search. With Jack AI operating on AWS, a lot of this work is now dealt with robotically. Journey pictures and brief notes uploaded by guides are transformed into a whole set of selling prepared property. This consists of journey studies, Search engine marketing wealthy web site content material, social captions, and electronic mail updates, produced in minutes somewhat than hours.

Jack AI adoption has steadily climbed, rising from simply over 100 studies in early 2025 to almost 340 studies by July 2025. This rise displays a broader shift in our guides and the out of doors trade, the place guides who as soon as hesitated to embrace know-how and digital advertising at the moment are counting on Jack AI to construct and develop their on-line presence.

Content material output has scaled dramatically, from beneath 800 property in early 2025 to greater than 2,500 by midsummer. Every journey report produces a number of deliverables—together with Search engine marketing artifacts for information web sites, captions for Instagram and Fb, and narrative descriptions tailor-made to electronic mail advertising—permitting guides to keep up an genuine, constant presence throughout channels.

Content material is delivered at peak relevance. With asset technology time dropping from 13 minutes in December 2024 to only two minutes by August 2025, Jack AI ensures that journey studies, social media posts, and electronic mail campaigns are prepared nearly instantly after a visit concludes. This pace permits guides to achieve purchasers and their networks whereas the expertise continues to be recent, driving larger engagement throughout social channels and quicker responses to post-trip emails. Automated emails showcasing latest journeys attain previous and present clients inside hours, serving to convert constructive shopper vitality into repeat bookings and word-of-mouth advertising.

Scaling income

The monetary affect of those enhancements has been clear. Among the many 5 most lively guides utilizing Jack AI:

• Common month-to-month income grew from roughly $3,000 in January 2025 to greater than $27,000 by July 2025—an almost 9× improve in simply six months.
• Guides credited the expansion to their capacity to keep up a gentle circulation of content material that boosted visibility in engines like google, drove engagement on social networks, and finally transformed into new bookings.

Conclusion

Maybe most significantly, guides have embraced Jack AI not solely as a reporting software, however as a core a part of operating and rising their companies. By automating journey report creation, Search engine marketing enhancements, social media content material, and electronic mail campaigns, Jack AI has turn into an integral a part of day by day operations, lowering the burden of selling whereas sustaining the authenticity of every information’s voice. Its capacity to determine species, estimate measurement, and incorporate real-world journey situations into content material provides a degree of element and engagement that guides and their purchasers worth. Jack AI delivers this performance reliably and constantly, dealing with rising volumes of media and journey knowledge with out interruption. The system’s serverless structure makes positive that as adoption continues to extend, efficiency stays excessive, and guides can give attention to what they do greatest: delivering distinctive out of doors experiences. These outcomes present how Jack AI helps out of doors guides get well hours spent on handbook content material creation, keep a constant on-line presence, and drive bookings. Constructed on scalable AWS infrastructure, the system turns a time-intensive process into an automatic, repeatable workflow.

Know extra about Amazon SageMaker AI to get began.

In regards to the authors

“David Lord – CEO”

“David” is a 3x Entrepreneur and Ernst & Younger Entrepreneur of the Yr. David is a lifelong fly fisherman with 25+ years’ expertise fishing with guides. Guidesly is a ardour undertaking that mixes David’s ardour for fly fishing, information experiences and being an entrepreneur.

“Taylor Lord – Head of Product”

“Taylor” is an skilled product and advertising chief. Taylor was an InsureTech / AI startup Hello Marley (Collection B) government earlier than becoming a member of Guidesly. Taylor has labored at two startups and is a wonderful product and advertising government.

“Shiva Prasad – Head of Know-how”

“Shiva” leads the know-how group in India. Shiva is a SaaS professional who has led improvement groups for 15 years. Shiva and David have labored collectively twice beforehand. That is his third startup, previously working for JumpStart Video games & RazorGator.

“Anup Banasavalli Hiriyanagowda – AI Affiliate/Information Scientist”

“Anup” is an AI Engineer at Guidesly, who enjoys constructing real-world AI programs. He makes a speciality of machine studying frameworks, massive language fashions, laptop imaginative and prescient, and scalable software program pipelines, utilizing AWS providers to design and deploy production-ready options.

“Nikhil Chandra – AI Affiliate”

“Nikhil” Naveen Chandra is an AI Affiliate Engineer at Guidesly, specializing in serverless AWS structure and AI-powered programs for large-scale communication and automation.