Introduction

The global construction industry was valued at US$13.7 trillion in 2021 and is expected to grow
to US$15.7 trillion by 2025. This growth is driven by several factors, including:

● Increasing urbanisation
● Rising disposable incomes
● Growing infrastructure needs
● A shift towards sustainable construction practices

The Indian construction market is estimated to be worth around $700 billion in 2022 and is
projected to reach $1.4 trillion by 2025, with an annual compound growth rate (CAGR) of over
5%. This growth is expected to be fueled by continued urbanisation, infrastructure development,
and a growing middle class.

Key Sectors:

The Indian construction industry comprises various sub-sectors, including:

● Residential construction: Comprising over 60% of the industry's output, this sector is
primarily fueled by population growth and urbanisation.
● Infrastructure construction: The government has set an ambitious target of
constructing roads worth INR 15 lakh crore ($212.80 billion) over the next two years, and
allocated a substantial INR 1,69,637 crore ($24.27 billion) in the Union Budget 2020-21
for transport infrastructure development.
● Commercial construction: Meeting the increasing demand for office buildings,
shopping malls, and hotels, this sector is supported by the expanding business
landscape and overall economic growth.
● Industrial construction: Driven by manufacturing and industrial development, this
sub-sector plays a pivotal role in diversifying the Indian construction industry.

Sector Market Size (2022) CAGR (2022-2025)

Residential construction $420 billion 6%

Infrastructure construction $180 billion 8%

Commercial construction $60 billion 7%

Industrial construction $40 billion 5%

The Problem
The construction industry faces significant productivity challenges. A single day's delay leads to
substantial cost overruns given the labour-intensive nature of the industry. Real-time visibility of
workers, machinery, and job sites is crucial to make better decisions to mitigate these costs.
Additionally, frequent reworks result from miscommunication and inadequate site visibility,
further impacting project efficiency.

Existing challenges across industry (based on customer discovery)

● Lack of granular insights during execution for decision making

● Miscommunication leading to delays in progress
● Lack of documentation and learning from past mistakes

Critical Components

● Progress Tracking:
○ Comparing 3D models (as-built) with 2D/3D designs.
○ Tracking construction progress with project schedules (Primavera/MS Project).
● Inventory Management:
○ Track real time usage of manpower, heavy machinery and raw materials.
● Documentation:
○ Creating project reports tailored to customer requirements.
○ Documenting all snags, issues and delays for future references.
○ Documenting site execution progress for stakeholder accountability

How does it affect a company?

Time and Cost Overruns eat into the already slim margins for large contractors. Each day of
throughout lost due to mismanagement and miscommunication is a significant financial loss to
the company.

Some alarming statistics about these overruns:

● Infrastructure projects face on average 50% time overruns with the actual average delay
~ 20 months.
● An average typical site (100 foremen, 3 planning engineers, 1 planning manager, 1
project manager) costs ~30L INR every month to function (just considering the
opportunity cost of the manpower and their salaries, not even considering other variables
like machines, equipment, electricity, water, etc.)
● Thus, every day of delay for such a site causes losses of ~ 1L INR to the company
● On top of this, every project has a delay penalty of ~0.1-0.5% of total project value per
week (typically capped at 10% aggregate)
According to a study conducted by KPMG on infrastructure projects in the country, ~20% of
these delays are caused by miscommunication, large cycle times of communication, ad-hoc
planning and a reactive approach to snags rather than a predictive approach.

Even though this gives us a reasonable estimate of how much these problems typically cost a
company, a better way is to narrow down and focus on their individual repercussions.

On large sites, such as linear projects like highways and metros, multiple sub-sites exist,
creating data "silos" that result in communication lags and blind spots between these sub-sites.
These challenges fall within the purview of the planning and project management teams.
Consequently, these teams become overburdened as the process of collecting, consolidating,
and reformatting multiple Daily Progress Reports (DPRs) consumes a significant portion of their
day (approximately 4 hours). This, in turn, relegates important tasks like snag documentation for
future reference and predictive planning for upcoming activities to the backseat, with these tasks
being addressed only once a month or so. As a consequence, numerous opportunities to
minimise delays and costs are missed.

For a planning team handling 100 workers, there is a potential to save 1.25L every month.
Moreover, with every month of delay minimised, an additional 30 lakh INR can be reclaimed, in
addition to cost reductions related to equipment rentals and other operational expenditures. This
can be done through better data collection, consolidation and communication from multiple
fronts across sub-sites across stakeholders without any human involvement.

Current ways of tackling it and why they don’t work

Operational methods

● Construction site engineers submit Daily Progress Reports (DPRs) for their specific
tasks. This DPR consists of the relevant progress details of the work package, as well as
details about the manpower, inventory and machines used for that day. The planning
engineer then links each DPR to the relevant work packet and updates two project
management tools, one for client progress monitoring and another for internal company
use.
● Planning managers typically look after ~30-40 site engineers in a large site. They
consolidate information from all these DPRs and contextualise the information to the
working schedule that is maintained usually in Excel Sheets. They then prepare DPRs
for stakeholders higher up in the chain.
● Due to the manual nature of information collection and processing, current update
meetings are the only source of information for most stakeholders. This is much much
less granular than expected.
● When using a BIM (Building Information Modeling) tool, a dedicated 4D modeller is hired
to incorporate progress data into the BIM. Each 4D modeller costs nearly 50,000 INR per
 month, and on average, each construction site requires four such engineers solely for
data input into the project management application.

Existing non-AI digital solutions (OnSite, PowerPlay, Asite, Procore, etc.)

● This category of solutions contain all features like stakeholder communication, snag
management and schedule tracking of as-built vs as-planned. But they face a major
hurdle in the Indian market. They all rely on site personnel to manually update and feed
in all the data. This causes major problems and has led to very sparse adoption.
● Some of these tools also seek to completely replace some existing legacy workflows.
This is not at all appreciated in an industry where those analogue workflows have
evolved over decades to become the most optimal way to do things. A complete
upending of these workflows also causes the companies time and effort to get used to.
Instead, the more pragmatic approach is to see where these legacy workflows are failing
or inefficient, and then solve those.

Existing AI based digital solutions (Openspace, Doxel, Siteaware, etc.)

● Almost all these solutions use Computer Vision to detect on-site processes and track
progress and extract useful information. This eliminates the manual nature of data
collection and insight extraction that plagued the solutions above.
● But, most of these solutions were built for the US construction industry, and hence have
been trained on datasets from the sites in the US. Fundamentally, western construction
methods heavily rely on drywall, while Indian and other asian countries use brick and
cement. This means that these solutions cannot be directly utilised here.

Why now? Why us?

The construction Industry in India is expected to reach $1.4 Tn by 2025.

● The pandemic taught the world that virtual ways of working and monitoring are feasible.
Multiple General contractors have started adopting digital solutions to automate parts of
their workflows
● First generation Construction Tech startups have shifted industry perception towards the
use of technology in this sector. Company now consider such solutions a serious
investment, rather than just a marketing gimmick
● There is a growing shortage of workforce and labour in all countries. This has also paved
the way for more technologies that automate manual and laborious workflows
● Global economic slowdown has compelled companies to prioritize addressing existing
inefficiencies in order to enhance their slim profit margins, especially in growing nations
● The AI hype taking over the world in the last 5 years is finally catching up to this legacy
industry, being a lagging adopter of technological trends. Most CXOs we talked to
believe that some form of AI solutions will boost their workflows
The founders have complimentary technical skills, as well as experience building in the
construction tech industry. Harshil’s strengths lie in robotics and building hardware products,
with past experiences being the Autonomy Lead at a medical drone delivery startup. Kanao’s
(Sid) strengths lie in building enterprise AI products for the construction industry. She also
founded an eVTOL startup in her undergraduate days.

Solution
We’re building a tool that helps decision makers at various stages in the stakeholder chain take
better and more informed decisions with the power of reliable insights from high quality and
repeatable visual site data.

Why visual data? Because it provides the richest form of context there is. Current ways to do
this are by walking the site and looking at different areas. We seek to provide the same context,
but to all stakeholders without having to walk the site every time.

Any solution that provides insights derived from data is inherently constrained in performance by
the quality of the data it gets. Hence, high quality data becomes a very important piece of the
puzzle.

Precursor - Dataset and Training

We’ve developed a generalisable model that can identify processes and context from Indian
construction sites. But this model needs to be finetuned on visual data for every company, to get
absolutely accurate performance, as there are process and semantic differences from company
to company.
Hence, we start by procuring a dataset from the company (CCTV feeds, helmet mounted cam
footage, etc.). In case the company does not have a dataset, we create our high quality own
dataset of processes, that we then use to finetune our AI model. This ensures we achieve
benchmark performance on every site!

Ofcourse, all datasets collected are protected by NDAs and privacy safeguards we have in
place against sharing that data externally.

Data collection
● Initially, for the pilots, we plan to utilise existing CCTV grids, for the ease of good quality
data collection. But moving forward, this will run into coverage issues.
● Hence, we plan to move to helmet mounted cameras and eventually, fully autonomous
UAVs that can capture reliable, high quality, high coverage data repeatedly.
We’ve used AI to replicate the way humans inherently look and process data. Firstly, context of
the visible process is identified (What is the process that is going on here? How does that fit into
what our goal is?)
Then, context about the extent of work is extracted. Progress is quantified and correlated to
parts of the schedule. Current progress is used to predict future clashes and take informed
decisions about asset and resource allocation.
This solution is generalizable across different kinds of sites, as visual context regarding
execution enables visibility, accountability and faster cycle times for snag management. When
snags are reliably identified faster, stakeholders can manage them better without any
finger-pointing. It frees us the decision makers’ time by letting them focus on more pragmatic
future planning rather than worrying about getting a reliable and holistic picture of the site.

Data Analysis and documentation

This consists broadly of 3 modules:

● Context Extraction
○ This module identifies ongoing activities by analyzing on-site elements like
materials, manpower, and machinery data, using a fine-tuned deep learning
model to recognize processes such as rebars attachment, shuttering, and
concrete pouring.
○ It correlates the observed object with its corresponding event in the schedule and
the specific section of the 2D drawings / 3D models, given an accurate tentative
object position. For this, it uses Volumetric Comparison of captured 3D model
and designs.
○ Site Daily Progress Reports (DPRs) are employed to validate the executed work
package.
○ Output : Identification of each ongoing process in incoming visual data, and its
identification in the planned schedule (Primavera or MsProject)

● Schedule Tracking
○ Granular progress is updated and shared with stakeholders by updating the
project's scheduling software (Primavera, MS Project, Excel) once available.
○ Output: Auto updating the progress on the scheduling software without any
human intervention. Potential clashes, causes of delays, and critical activities
getting affected by this progress update are highlighted.

● Quantity Estimation
○ Estimation of material, manpower, and machinery usage is performed, and the
relevant data is maintained.
○ Site Daily Progress Reports (DPRs) are also used to enhance and validate these
estimates.
○ Output: Estimated quantity of material used, manpower used and machinery
used.
Who is this solution for?
General Contractors: Key decision makers within the stakeholder chain that handle design,
execution and procurement related domains. Faster, high quality and reliable data enables them
to take key decisions that save delays and costs. Reduced cycle times of snag management
and no finger-pointing for issues. At no extra operation manpower effort.

Customers: Project teams that outsource and subcontract execution. Better visibility over the
project. Reduced cycle times of snag management and no finger-pointing for issues. At no extra
operation manpower effort.